/[emacs]/emacs/etc/PROBLEMS
ViewVC logotype

Diff of /emacs/etc/PROBLEMS

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1.166 by tamm, Sun May 30 00:35:44 2004 UTC revision 1.166.2.1 by miles, Sat Sep 4 12:01:05 2004 UTC
# Line 1  Line 1 
1  This file describes various problems that have been encountered  This file describes various problems that have been encountered
2  in compiling, installing and running GNU Emacs.  in compiling, installing and running GNU Emacs.  Try doing Ctl t
3    and browsing through the outline headers.
4    
5  * Environment Variables from dotfiles are ignored with Mac OS X (Carbon).  * Emacs startup failures
6    
7  When starting Emacs from the Dock or the Finder on Mac OS X, the  ** Emacs fails to start, complaining about missing fonts.
 environment variables that are set up in dotfiles, such as .cshrc or  
 .profile, are ignored.  This is because the Finder and Dock are not  
 started from a shell, but instead from the Window Manager itself.  
   
 The workaround for this is to create a .MacOSX/environment.plist file to  
 setup these environment variables.  These environment variables will  
 apply to all processes regardless of where they are started.  
 For me information, see http://developer.apple.com/qa/qa2001/qa1067.html.  
   
 * Segfault on GNU/Linux using certain recent versions of the Linux kernel.  
   
 With certain recent Linux kernels (like the one of Redhat Fedora Core  
 1), the new "Exec-shield" functionality is enabled by default, which  
 creates a different memory layout that breaks the emacs dumper.  
   
 You can check the Exec-shield state like this:  
   
     cat /proc/sys/kernel/exec-shield  
   
 It returns 1 or 2 when Exec-shield is enabled, 0 otherwise.  Please  
 read your system documentation for more details on Exec-shield and  
 associated commands.  
   
 When Exec-shield is enabled, building Emacs will segfault during the  
 execution of this command:  
   
 temacs --batch --load loadup [dump|bootstrap]  
   
 To work around this problem, it is necessary to temporarily disable  
 Exec-shield while building Emacs, using the `setarch' command like  
 this:  
   
     setarch i386 ./configure <configure parameters>  
     setarch i386 make <make parameters>  
   
 * Characters are displayed as empty boxes or with wrong font under X.  
   
 This can occur when two different versions of FontConfig are used.  
 For example, XFree86 4.3.0 has one version and Gnome usually comes  
 with a newer version.  Emacs compiled with --with-gtk will then use  
 the newer version.  In most cases the problem can be temporarily  
 fixed by stopping the application that has the error (it can be  
 Emacs or any other application), removing ~/.fonts.cache-1,  
 and then start the application again.  
 If removing ~/.fonts.cache-1 and restarting doesn't help, the  
 application with problem must be recompiled with the same version  
 of FontConfig as the rest of the system uses.  For KDE, it is  
 sufficient to recompile Qt.  
   
 * Process output truncated on Mac OS X (Carbon) when using pty's.  
   
 There appears to be a problem with the implementation of pty's on the  
 Mac OS X that causes process output to be truncated.  To avoid this,  
 leave process-connection-type set to its default value of nil.  
   
 * Emacs crashes on Mac OS X (Carbon) after system software upgrade.  
   
 This problem seems to be now solved by Steven Tamm's patch to  
 unexmacosx.c on Nov 24, 2002.  
   
 Between Mac OS X release 10.2.1 and 10.2.2 there was an incompatible  
 change in the memory allocator that causes a EXC_BAD_ACCESS error near  
 xrealloc().  Relinking the application (by deleting src/temacs and  
 running make) will solve the problem.  It appears to be caused by some  
 problems with the unexec code and its interaction with libSystem.B.  
   
 * Emacs crashes with SIGSEGV in XtInitializeWidgetClass  
   
 It crashes on X, but runs fine when called with option "-nw".  
   
 This has been observed when Emacs is linked with GNU ld but without passing  
 the -z nocombreloc flag.  Emacs normally knows to pass the -z nocombreloc  
 flag when needed, so if you come across a situation where the flag is  
 necessary but missing, please report it via M-x report-emacs-bug.  
   
 On platforms such as Solaris, you can also work around this problem by  
 configuring your compiler to use the native linker instead of GNU ld.  
           
 * Characters from the mule-unicode charsets aren't displayed under X.  
8    
9  XFree86 4 contains many fonts in iso10646-1 encoding which have  A typical error message might be something like
 minimal character repertoires (whereas the encoding part of the font  
 name is meant to be a reasonable indication of the repertoire  
 according to the XLFD spec).  Emacs may choose one of these to display  
 characters from the mule-unicode charsets and then typically won't be  
 able to find the glyphs to display many characters.  (Check with C-u  
 C-x = .)  To avoid this, you may need to use a fontset which sets the  
 font for the mule-unicode sets explicitly.  E.g. to use GNU unifont,  
 include in the fontset spec:  
   
 mule-unicode-2500-33ff:-gnu-unifont-*-iso10646-1,\  
 mule-unicode-e000-ffff:-gnu-unifont-*-iso10646-1,\  
 mule-unicode-0100-24ff:-gnu-unifont-*-iso10646-1  
10    
11  * The UTF-8/16/7 coding systems don't encode CJK (Far Eastern) characters.    No fonts match `-*-fixed-medium-r-*--6-*-*-*-*-*-iso8859-1'
12    
13  Emacs by default only supports the parts of the Unicode BMP whose code  This happens because some X resource specifies a bad font family for
14  points are in the ranges 0000-33ff and e000-ffff.  This excludes: most  Emacs to use.  The possible places where this specification might be
15  of CJK, Yi and Hangul, as well as everything outside the BMP.  are:
16    
17  If you read UTF-8 data with code points outside these ranges, the    - in your ~/.Xdefaults file
 characters appear in the buffer as raw bytes of the original UTF-8  
 (composed into a single quasi-character) and they will be written back  
 correctly as UTF-8, assuming you don't break the composed sequences.  
 If you read such characters from UTF-16 or UTF-7 data, they are  
 substituted with the Unicode `replacement character', and you lose  
 information.  
18    
19  To edit such UTF data, turn on Utf-Translate-Cjk mode, which makes    - client-side X resource file, such as  ~/Emacs or
20  many common CJK characters available for encoding and decoding and can      /usr/X11R6/lib/app-defaults/Emacs or
21  be extended by updating the tables it uses.  This also allows you to      /usr/X11R6/lib/X11/app-defaults/Emacs
 save as UTF buffers containing characters decoded by the chinese-,  
 japanese- and korean- coding systems, e.g. cut and pasted from  
 elsewhere.  
22    
23  * Problems with file dialogs in Emacs built with Open Motif.  One of these files might have bad or malformed specification of a
24    fontset that Emacs should use.  To fix the problem, you need to find
25    the problematic line(s) and correct them.
26    
27  When Emacs 21 is built with Open Motif 2.1, it can happen that the  ** Emacs aborts while starting up, only when run without X.
 graphical file dialog boxes do not work properly.  The "OK", "Filter"  
 and "Cancel" buttons do not respond to mouse clicks.  Dragging the  
 file dialog window usually causes the buttons to work again.  
28    
29  The solution is to use LessTif instead.  LessTif is a free replacement  This problem often results from compiling Emacs with GCC when GCC was
30  for Motif.  See the file INSTALL for information on how to do this.  installed incorrectly.  The usual error in installing GCC is to
31    specify --includedir=/usr/include.  Installation of GCC makes
32    corrected copies of the system header files.  GCC is supposed to use
33    the corrected copies in preference to the original system headers.
34    Specifying --includedir=/usr/include causes the original system header
35    files to be used.  On some systems, the definition of ioctl in the
36    original system header files is invalid for ANSI C and causes Emacs
37    not to work.
38    
39  Another workaround is not to use the mouse to trigger file prompts,  The fix is to reinstall GCC, and this time do not specify --includedir
40  but to use the keyboard.  This way, you will be prompted for a file in  when you configure it.  Then recompile Emacs.  Specifying --includedir
41  the minibuffer instead of a graphical file dialog.  is appropriate only in very special cases and it should *never* be the
42    same directory where system header files are kept.
43    
44  * Emacs reports a BadAtom error (from X) running on Solaris 7 or 8.  ** Emacs does not start, complaining that it cannot open termcap database file.
45    
46  This happens when Emacs was built on some other version of Solaris.  If your system uses Terminfo rather than termcap (most modern
47  Rebuild it on Solaris 8.  systems do), this could happen if the proper version of
48    ncurses is not visible to the Emacs configure script (i.e. it
49    cannot be found along the usual path the linker looks for
50    libraries). It can happen because your version of ncurses is
51    obsolete, or is available only in form of binaries.
52    
53  * Mule-UCS loads very slowly.  The solution is to install an up-to-date version of ncurses in
54    the developer's form (header files, static libraries and
55    symbolic links); in some GNU/Linux distributions (e.g. Debian)
56    it constitutes a separate package.
57    
58  Changes to Emacs internals interact badly with Mule-UCS's `un-define'  ** Emacs 20 and later fails to load Lisp files at startup.
 library, which is the usual interface to Mule-UCS.  Apply the  
 following patch to Mule-UCS 0.84 and rebuild it.  That will help,  
 though loading will still be slower than in Emacs 20.  (Some  
 distributions, such as Debian, may already have applied such a patch.)  
59    
60  --- lisp/un-define.el   6 Mar 2001 22:41:38 -0000       1.30  The typical error message might be like this:
 +++ lisp/un-define.el   19 Apr 2002 18:34:26 -0000  
 @@ -610,13 +624,21 @@ by calling post-read-conversion and pre-  
61    
62    (mapcar    "Cannot open load file: fontset"
    (lambda (x)  
 -    (mapcar  
 -     (lambda (y)  
 -       (mucs-define-coding-system  
 -       (nth 0 y) (nth 1 y) (nth 2 y)  
 -       (nth 3 y) (nth 4 y) (nth 5 y) (nth 6 y))  
 -       (coding-system-put (car y) 'alias-coding-systems (list (car x))))  
 -     (cdr x)))  
 +    (if (fboundp 'register-char-codings)  
 +       ;; Mule 5, where we don't need the eol-type specified and  
 +       ;; register-char-codings may be very slow for these coding  
 +       ;; system definitions.  
 +       (let ((y (cadr x)))  
 +         (mucs-define-coding-system  
 +          (car x) (nth 1 y) (nth 2 y)  
 +          (nth 3 y) (nth 4 y) (nth 5 y)))  
 +      (mapcar  
 +       (lambda (y)  
 +        (mucs-define-coding-system  
 +         (nth 0 y) (nth 1 y) (nth 2 y)  
 +         (nth 3 y) (nth 4 y) (nth 5 y) (nth 6 y))  
 +        (coding-system-put (car y) 'alias-coding-systems (list (car x)))))  
 +      (cdr x)))  
    `((utf-8  
       (utf-8-unix  
        ?u "UTF-8 coding system"  
63    
64  Note that Emacs has native support for Unicode, roughly equivalent to  This could happen if you compress the file lisp/subdirs.el.  That file
65  Mule-UCS's, so you may not need it.  tells Emacs what are the directories where it should look for Lisp
66    files.  Emacs cannot work with subdirs.el compressed, since the
67    Auto-compress mode it needs for this will not be loaded until later,
68    when your .emacs file is processed.  (The package `fontset.el' is
69    required to set up fonts used to display text on window systems, and
70    it's loaded very early in the startup procedure.)
71    
72  * Building Emacs with GCC 2.9x fails in the `src' directory.  Similarly, any other .el file for which there's no corresponding .elc
73    file could fail to load if it is compressed.
74    
75  This may happen if you use a development version of GNU `cpp' from one  The solution is to uncompress all .el files which don't have a .elc
76  of the GCC snapshots between Oct 2000 and Feb 2001, or from a released  file.
 version of GCC newer than 2.95.2 which was prepared around those  
 dates; similar problems were reported with some snapshots of GCC 3.1  
 around Sep 30 2001.  The preprocessor in those versions is  
 incompatible with a traditional Unix cpp (e.g., it expands ".." into  
 ". .", which breaks relative file names that reference the parent  
 directory; or inserts TAB characters before lines that set Make  
 variables).  
77    
78  The solution is to make sure the preprocessor is run with the  Another possible reason for such failures is stale *.elc files
79  `-traditional' option.  The `configure' script does that automatically  lurking somewhere on your load-path.  The following command will
80  when it detects the known problems in your cpp, but you might hit some  print any duplicate Lisp files that are present in load-path:
 unknown ones.  To force the `configure' script to use `-traditional',  
 run the script like this:  
81    
82    CPP='gcc -E -traditional' ./configure ...      emacs -q -batch -f list-load-path-shadows
83    
84  (replace the ellipsis "..." with any additional arguments you pass to  If this command prints any file names, some of these files are stale,
85  the script).  and should be deleted or their directories removed from your
86    load-path.
87    
88  Note that this problem does not pertain to the MS-Windows port of  ** Emacs prints an error at startup after upgrading from an earlier version.
 Emacs, since it doesn't use the preprocessor to generate Makefiles.  
89    
90  * Building Emacs with a system compiler fails to link because of an  An example of such an error is:
 undefined symbol such as __eprintf which does not appear in Emacs.  
91    
92  This can happen if some of the libraries linked into Emacs were built    x-complement-fontset-spec: "Wrong type argument: stringp, nil"
 with GCC, but Emacs itself is being linked with a compiler other than  
 GCC.  Object files compiled with GCC might need some helper functions  
 from libgcc.a, the library which comes with GCC, but the system  
 compiler does not instruct the linker to search libgcc.a during the  
 link stage.  
93    
94  A solution is to link with GCC, like this:  This can be another symptom of stale *.elc files in your load-path.
95    The following command will print any duplicate Lisp files that are
96    present in load-path:
97    
98          make CC=gcc      emacs -q -batch -f list-load-path-shadows
99    
100  Since the .o object files already exist, this will not recompile Emacs  If this command prints any file names, some of these files are stale,
101  with GCC, but just restart by trying again to link temacs.  and should be deleted or their directories removed from your
102    load-path.
103    
104  * Building the MS-Windows port with Cygwin GCC can fail.  ** With X11R6.4, public-patch-3, Emacs crashes at startup.
105    
106  Emacs may not build using recent Cygwin builds of GCC, such as Cygwin  Reportedly this patch in X fixes the problem.
 version 1.1.8, using the default configure settings.  It appears to be  
 necessary to specify the -mwin32 flag when compiling, and define  
 __MSVCRT__, like so:  
107    
108    configure --with-gcc --cflags -mwin32 --cflags -D__MSVCRT__      --- xc/lib/X11/imInt.c~     Wed Jun 30 13:31:56 1999
109        +++ xc/lib/X11/imInt.c      Thu Jul  1 15:10:27 1999
110        @@ -1,4 +1,4 @@
111        -/* $TOG: imInt.c /main/5 1998/05/30 21:11:16 kaleb $ */
112        +/* $TOG: imInt.c /main/5 1998/05/30 21:11:16 kaleb $  */
113         /******************************************************************
114    
115  * Building the MS-Windows port fails with a CreateProcess failure.                  Copyright 1992, 1993, 1994 by FUJITSU LIMITED
116        @@ -166,8 +166,8 @@
117         _XimMakeImName(lcd)
118             XLCd      lcd;
119         {
120        -    char* begin;
121        -    char* end;
122        +    char* begin = NULL;
123        +    char* end = NULL;
124             char* ret;
125             int    i = 0;
126             char* ximmodifier = XIMMODIFIER;
127        @@ -182,7 +182,11 @@
128             }
129             ret = Xmalloc(end - begin + 2);
130             if (ret != NULL) {
131        -           (void)strncpy(ret, begin, end - begin + 1);
132        +   if (begin != NULL) {
133        +             (void)strncpy(ret, begin, end - begin + 1);
134        +        } else {
135        +     ret[0] = '\0';
136        +   }
137                ret[end - begin + 1] = '\0';
138             }
139             return ret;
140    
141  Some versions of mingw32 make on some versions of Windows do not seem  * Crash bugs
 to detect the shell correctly. Try "make SHELL=cmd.exe", or if that  
 fails, try running make from Cygwin bash instead.  
142    
143  * Building the MS-Windows port with Leim fails in the `leim' directory.  ** Emacs crashes in x-popup-dialog.
144    
145  The error message might be something like this:  This can happen if the dialog widget cannot find the font it wants to
146    use.  You can work around the problem by specifying another font with
147    an X resource--for example, `Emacs.dialog*.font: 9x15' (or any font that
148    happens to exist on your X server).
149    
150   Converting d:/emacs-21.3/leim/CXTERM-DIC/4Corner.tit to quail-package...  ** Emacs crashes when you use Bibtex mode.
  Invalid ENCODE: value in TIT dictionary  
  NMAKE : fatal error U1077: '"../src/obj-spd/i386/emacs.exe"' : return code  
        '0xffffffff'  
  Stop.  
151    
152  This can happen if the Leim distribution is unpacked with a program  This happens if your system puts a small limit on stack size.  You can
153  which converts the `*.tit' files to DOS-style CR-LF text format.  The  prevent the problem by using a suitable shell command (often `ulimit')
154  `*.tit' files in the leim/CXTERM-DIC directory require Unix-style line  to raise the stack size limit before you run Emacs.
 endings to compile properly, because Emacs reads them without any code  
 or EOL conversions.  
155    
156  The solution is to make sure the program used to unpack Leim does not  Patches to raise the stack size limit automatically in `main'
157  change the files' line endings behind your back.  The GNU FTP site has  (src/emacs.c) on various systems would be greatly appreciated.
 in the `/gnu/emacs/windows' directory a program called `djtarnt.exe'  
 which can be used to unpack `.tar.gz' and `.zip' archives without  
 mangling them.  
158    
159  * Emacs crashes when dumping itself on Mac PPC running Yellow Dog GNU/Linux.  ** Emacs crashes with SIGBUS or SIGSEGV on HPUX 9 after you delete a frame.
160    
161  The crashes happen inside the function Fmake_symbol; here's a typical  We think this is due to a bug in the X libraries provided by HP.  With
162  C backtrace printed by GDB:  the alternative X libraries in /usr/contrib/mitX11R5/lib, the problem
163    does not happen.
164    
165    0x190c0c0 in Fmake_symbol ()  ** Emacs crashes with SIGBUS or SIGSEGV on Solaris after you delete a frame.
   (gdb) where  
   #0  0x190c0c0 in Fmake_symbol ()  
   #1  0x1942ca4 in init_obarray ()  
   #2  0x18b3500 in main ()  
   #3  0x114371c in __libc_start_main (argc=5, argv=0x7ffff5b4, envp=0x7ffff5cc,  
166    
167  This could happen because GCC version 2.95 and later changed the base  We suspect that this is a similar bug in the X libraries provided by
168  of the load address to 0x10000000.  Emacs needs to be told about this,  Sun.  There is a report that one of these patches fixes the bug and
169  but we currently cannot do that automatically, because that breaks  makes the problem stop:
 other versions of GNU/Linux on the MacPPC.  Until we find a way to  
 distinguish between the Yellow Dog and the other varieties of  
 GNU/Linux systems on the PPC, you will have to manually uncomment the  
 following section near the end of the file src/m/macppc.h in the Emacs  
 distribution:  
170    
171    #if 0  /* This breaks things on PPC GNU/Linux except for Yellowdog,  105216-01 105393-01 105518-01 105621-01 105665-01 105615-02 105216-02
172              even with identical GCC, as, ld.  Let's take it out until we  105667-01 105401-08 105615-03 105621-02 105686-02 105736-01 105755-03
173              know what's really going on here.  */  106033-01 105379-01 105786-01 105181-04 105379-03 105786-04 105845-01
174    /* GCC 2.95 and newer on GNU/Linux PPC changed the load address to  105284-05 105669-02 105837-01 105837-02 105558-01 106125-02 105407-01
      0x10000000.  */  
   #if defined __linux__  
   #if __GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)  
   #define DATA_SEG_BITS  0x10000000  
   #endif  
   #endif  
   #endif /* 0 */  
175    
176  Remove the "#if 0" and "#endif" directives which surround this, save  Another person using a newer system (kernel patch level Generic_105181-06)
177  the file, and then reconfigure and rebuild Emacs.  The dumping process  suspects that the bug was fixed by one of these more recent patches:
 should now succeed.  
178    
179  * JPEG images aren't displayed.  106040-07  SunOS 5.6: X Input & Output Method patch
180    106222-01  OpenWindows 3.6: filemgr (ff.core) fixes
181    105284-12  Motif 1.2.7: sparc Runtime library patch
182    
183  This has been reported when Emacs is built with jpeg-6a library.  ** Error message `Symbol's value as variable is void: x', followed by
184  Upgrading to jpeg-6b solves the problem.  Configure checks for the  a segmentation fault and core dump.
 correct version, but this problem could occur if a binary built  
 against a shared libjpeg is run on a system with an older version.  
185    
186  * Building `ctags' for MS-Windows with the MinGW port of GCC fails.  This has been tracked to a bug in tar!  People report that tar erroneously
187    added a line like this at the beginning of files of Lisp code:
188    
189  This might happen due to a bug in the MinGW header assert.h, which     x FILENAME, N bytes, B tape blocks
 defines the `assert' macro with a trailing semi-colon.  The following  
 patch to assert.h should solve this:  
190    
191  *** include/assert.h.orig       Sun Nov  7 02:41:36 1999  If your tar has this problem, install GNU tar--if you can manage to
192  --- include/assert.h    Mon Jan 29 11:49:10 2001  untar it :-).
 ***************  
 *** 41,47 ****  
   /*  
    * If not debugging, assert does nothing.  
    */  
 ! #define assert(x)     ((void)0);  
193    
194    #else /* debugging enabled */  ** Crashes when displaying GIF images in Emacs built with version
195    libungif-4.1.0 are resolved by using version libungif-4.1.0b1.
196    Configure checks for the correct version, but this problem could occur
197    if a binary built against a shared libungif is run on a system with an
198    older version.
199    
200  --- 41,47 ----  ** Emacs aborts inside the function `tparam1'.
   /*  
    * If not debugging, assert does nothing.  
    */  
 ! #define assert(x)     ((void)0)  
201    
202    #else /* debugging enabled */  This can happen if Emacs was built without terminfo support, but the
203    terminal's capabilities use format that is only supported by terminfo.
204    If your system has ncurses installed, this might happen if your
205    version of ncurses is broken; upgrading to a newer version of ncurses
206    and reconfiguring and rebuilding Emacs should solve this.
207    
208    All modern systems support terminfo, so even if ncurses is not the
209    problem, you should look for a way to configure Emacs so that it uses
210    terminfo when built.
211    
212    ** Emacs crashes when using the Exceed 6.0 X server.
213    
214  * Improving performance with slow X connections  If you are using Exceed 6.1, upgrade to a later version.  This was
215    reported to prevent the crashes.
216    
217  There are several ways to improve this performance, any subset of which can  ** Emacs crashes with SIGSEGV in XtInitializeWidgetClass.
 be carried out at the same time:  
218    
219  1) If you don't need X Input Methods (XIM) for entering text in some  It crashes on X, but runs fine when called with option "-nw".
    language you use, you can improve performance on WAN links by using  
    the X resource useXIM to turn off use of XIM.  This does not affect  
    the use of Emacs' own input methods, which are part of the Leim  
    package.  
220    
221  2) If the connection is very slow, you might also want to consider  This has been observed when Emacs is linked with GNU ld but without passing
222     switching off scroll bars, menu bar, and tool bar.  the -z nocombreloc flag.  Emacs normally knows to pass the -z nocombreloc
223    flag when needed, so if you come across a situation where the flag is
224    necessary but missing, please report it via M-x report-emacs-bug.
225    
226  3) Use ssh to forward the X connection, and enable compression on this  On platforms such as Solaris, you can also work around this problem by
227     forwarded X connection (ssh -XC remotehostname emacs ...).  configuring your compiler to use the native linker instead of GNU ld.
228    
229  4) Use lbxproxy on the remote end of the connection.  This is an interface  * General runtime problems
    to the low bandwidth X extension in most modern X servers, which  
    improves performance dramatically, at the slight expense of correctness  
    of the X protocol.  lbxproxy acheives the performance gain by grouping  
    several X requests in one TCP packet and sending them off together,  
    instead of requiring a round-trip for each X request in a seperate  
    packet.  The switches that seem to work best for emacs are:  
     -noatomsfile  -nowinattr  -cheaterrors -cheatevents  
    Note that the -nograbcmap option is known to cause problems.  
    For more about lbxproxy, see:  
    http://www.xfree86.org/4.3.0/lbxproxy.1.html  
230    
231  * Getting a Meta key on the FreeBSD console  ** Lisp problems
232    
233  By default, neither Alt nor any other key acts as a Meta key on  *** Changes made to .el files do not take effect.
 FreeBSD, but this can be changed using kbdcontrol(1).  Dump the  
 current keymap to a file with the command  
234    
235    $ kbdcontrol -d >emacs.kbd  You may have forgotten to recompile them into .elc files.
236    Then the old .elc files will be loaded, and your changes
237    will not be seen.  To fix this, do M-x byte-recompile-directory
238    and specify the directory that contains the Lisp files.
239    
240  Edit emacs.kbd, and give the key you want to be the Meta key the  Emacs should print a warning when loading a .elc file which is older
241  definition `meta'.  For instance, if your keyboard has a ``Windows''  than the corresponding .el file.
 key with scan code 105, change the line for scan code 105 in emacs.kbd  
 to look like this  
242    
243    105   meta   meta   meta   meta   meta   meta   meta   meta    O  *** Watch out for .emacs files and EMACSLOADPATH environment vars.
244    
245  to make the Windows key the Meta key.  Load the new keymap with  These control the actions of Emacs.
246    ~/.emacs is your Emacs init file.
247    EMACSLOADPATH overrides which directories the function
248    "load" will search.
249    
250    $ kbdcontrol -l emacs.kbd  If you observe strange problems, check for these and get rid
251    of them, then try again.
252    
253  * Emacs' xterm-mouse-mode doesn't work on the Gnome terminal.  *** Using epop3.el package causes Emacs to signal an error.
254    
255  A symptom of this bug is that double-clicks insert a control sequence  The error message might be something like this:
 into the buffer.  The reason this happens is an apparent  
 incompatibility of the Gnome terminal with Xterm, which also affects  
 other programs using the Xterm mouse interface.  A problem report has  
 been filed.  
256    
257  * Emacs pauses for several seconds when changing the default font    "Lisp nesting exceeds max-lisp-eval-depth"
258    
259  This has been reported for fvwm 2.2.5 and the window manager of KDE  This happens because epop3 redefines the function gethash, which is a
260  2.1.  The reason for the pause is Xt waiting for a ConfigureNotify  built-in primitive beginning with Emacs 21.1.  We don't have a patch
261  event from the window manager, which the window manager doesn't send.  for epop3 that fixes this, but perhaps a newer version of epop3
262  Xt stops waiting after a default timeout of usually 5 seconds.  corrects that.
263    
264  A workaround for this is to add something like  *** Buffers from `with-output-to-temp-buffer' get set up in Help mode.
265    
266  emacs.waitForWM: false  Changes in Emacs 20.4 to the hooks used by that function cause
267    problems for some packages, specifically BBDB.  See the function's
268    documentation for the hooks involved.  BBDB 2.00.06 fixes the problem.
269    
270  to your X resources.  Alternatively, add `(wait-for-wm . nil)' to a  *** The Hyperbole package causes *Help* buffers not to be displayed in
271  frame's parameter list, like this:  Help mode due to setting `temp-buffer-show-hook' rather than using
272    `add-hook'.  Using `(add-hook 'temp-buffer-show-hook
273    'help-mode-maybe)' after loading Hyperbole should fix this.
274    
275     (modify-frame-parameters nil '((wait-for-wm . nil)))  ** Keyboard problems
276    
277  (this should go into your `.emacs' file).  *** "Compose Character" key does strange things when used as a Meta key.
278    
279  * Underlines appear at the wrong position.  If you define one key to serve as both Meta and Compose Character, you
280    will get strange results.  In previous Emacs versions, this "worked"
281    in that the key acted as Meta--that's because the older Emacs versions
282    did not try to support Compose Character.  Now Emacs tries to do
283    character composition in the standard X way.  This means that you
284    must pick one meaning or the other for any given key.
285    
286  This is caused by fonts having a wrong UNDERLINE_POSITION property.  You can use both functions (Meta, and Compose Character) if you assign
287  Examples are the font 7x13 on XFree prior to version 4.1, or the jmk  them to two different keys.
 neep font from the Debian xfonts-jmk package.  To circumvent this  
 problem, set x-use-underline-position-properties to nil in your  
 `.emacs'.  
288    
289  To see what is the value of UNDERLINE_POSITION defined by the font,  *** C-z just refreshes the screen instead of suspending Emacs.
 type `xlsfonts -lll FONT' and look at the font's UNDERLINE_POSITION  
 property.  
290    
291  * When using Xaw3d scroll bars without arrows, the very first mouse  You are probably using a shell that doesn't support job control, even
292  click in a scroll bar might be ignored by the scroll bar widget.  This  though the system itself is capable of it.  Either use a different shell,
293  is probably a bug in Xaw3d; when Xaw3d is compiled with arrows, the  or set the variable `cannot-suspend' to a non-nil value.
 problem disappears.  
294    
295  * There are known binary incompatibilities between Xaw, Xaw3d, neXtaw,  *** With M-x enable-flow-control, you need to type C-\ twice
296  XawM and the few other derivatives of Xaw.  So when you compile with  to do incremental search--a single C-\ gets no response.
 one of these, it may not work to dynamically link with another one.  
 For example, strange problems, such as Emacs exiting when you type  
 "C-x 1", were reported when Emacs compiled with Xaw3d and libXaw was  
 used with neXtaw at run time.  
297    
298  The solution is to rebuild Emacs with the toolkit version you actually  This has been traced to communicating with your machine via kermit,
299  want to use, or set LD_PRELOAD to preload the same toolkit version you  with C-\ as the kermit escape character.  One solution is to use
300  built Emacs with.  another escape character in kermit.  One user did
301    
302  * Clicking C-mouse-2 in the scroll bar doesn't split the window.     set escape-character 17
303    
304  This currently doesn't work with scroll-bar widgets (and we don't know  in his .kermrc file, to make C-q the kermit escape character.
 a good way of implementing it with widgets).  If Emacs is configured  
 --without-toolkit-scroll-bars, C-mouse-2 on the scroll bar does work.  
305    
306  * Emacs aborts inside the function `tparam1'.  ** Mailers and other helper programs
307    
308  This can happen if Emacs was built without terminfo support, but the  *** movemail compiled with POP support can't connect to the POP server.
 terminal's capabilities use format that is only supported by terminfo.  
 If your system has ncurses installed, this might happen if your  
 version of ncurses is broken; upgrading to a newer version of ncurses  
 and reconfiguring and rebuilding Emacs should solve this.  
309    
310  All modern systems support terminfo, so even if ncurses is not the  Make sure that the `pop' entry in /etc/services, or in the services
311  problem, you should look for a way to configure Emacs so that it uses  NIS map if your machine uses NIS, has the same port number as the
312  terminfo when built.  entry on the POP server.  A common error is for the POP server to be
313    listening on port 110, the assigned port for the POP3 protocol, while
314    the client is trying to connect on port 109, the assigned port for the
315    old POP protocol.
316    
317  * Error messages about undefined colors on X.  *** RMAIL gets error getting new mail.
318    
319  The messages might say something like this:  RMAIL gets new mail from /usr/spool/mail/$USER using a program
320    called `movemail'.  This program interlocks with /bin/mail using
321    the protocol defined by /bin/mail.
322    
323     Unable to load color "grey95"  There are two different protocols in general use.  One of them uses
324    the `flock' system call.  The other involves creating a lock file;
325    `movemail' must be able to write in /usr/spool/mail in order to do
326    this.  You control which one is used by defining, or not defining,
327    the macro MAIL_USE_FLOCK in config.h or the m- or s- file it includes.
328    IF YOU DON'T USE THE FORM OF INTERLOCKING THAT IS NORMAL ON YOUR
329    SYSTEM, YOU CAN LOSE MAIL!
330    
331  (typically, in the `*Messages*' buffer), or something like this:  If your system uses the lock file protocol, and fascist restrictions
332    prevent ordinary users from writing the lock files in /usr/spool/mail,
333    you may need to make `movemail' setgid to a suitable group such as
334    `mail'.  You can use these commands (as root):
335    
336    Error while displaying tooltip: (error Undefined color lightyellow)          chgrp mail movemail
337            chmod 2755 movemail
338    
339  These problems could happen if some other X program has used up too  If your system uses the lock file protocol, and fascist restrictions
340  many colors of the X palette, leaving Emacs with insufficient system  prevent ordinary users from writing the lock files in /usr/spool/mail,
341  resources to load all the colors it needs.  you may need to make `movemail' setgid to a suitable group such as
342    `mail'.  To do this,  use the following commands (as root) after doing the
343    make install.
344    
345  A solution is to exit the offending X programs before starting Emacs.          chgrp mail movemail
346            chmod 2755 movemail
347    
348  * Colors are not available on a tty or in xterm.  Installation normally copies movemail from the build directory to an
349    installation directory which is usually under /usr/local/lib.  The
350    installed copy of movemail is usually in the directory
351    /usr/local/lib/emacs/VERSION/TARGET.  You must change the group and
352    mode of the installed copy; changing the group and mode of the build
353    directory copy is ineffective.
354    
355  Emacs 21 supports colors on character terminals and terminal  *** rcs2log gives you the awk error message "too many fields".
 emulators, but this support relies on the terminfo or termcap database  
 entry to specify that the display supports color.  Emacs looks at the  
 "Co" capability for the terminal to find out how many colors are  
 supported; it should be non-zero to activate the color support within  
 Emacs.  (Most color terminals support 8 or 16 colors.)  If your system  
 uses terminfo, the name of the capability equivalent to "Co" is  
 "colors".  
356    
357  In addition to the "Co" capability, Emacs needs the "op" (for  This is due to an arbitrary limit in certain versions of awk.
358  ``original pair'') capability, which tells how to switch the terminal  The solution is to use gawk (GNU awk).
 back to the default foreground and background colors.  Emacs will not  
 use colors if this capability is not defined.  If your terminal entry  
 doesn't provide such a capability, try using the ANSI standard escape  
 sequence \E[00m (that is, define a new termcap/terminfo entry and make  
 it use your current terminal's entry plus \E[00m for the "op"  
 capability).  
359    
360  Finally, the "NC" capability (terminfo name: "ncv") tells Emacs which  ** Problems with hostname resolution
 attributes cannot be used with colors.  Setting this capability  
 incorrectly might have the effect of disabling colors; try setting  
 this capability to `0' (zero) and see if that helps.  
361    
362  Emacs uses the database entry for the terminal whose name is the value  *** Emacs fails to understand most Internet host names, even though
363  of the environment variable TERM.  With `xterm', a common terminal  the names work properly with other programs on the same system.
364  entry that supports color is `xterm-color', so setting TERM's value to  *** Emacs won't work with X-windows if the value of DISPLAY is HOSTNAME:0.
365  `xterm-color' might activate the color support on an xterm-compatible  *** GNUs can't make contact with the specified host for nntp.
 emulator.  
366    
367  Beginning with version 21.4, Emacs supports the --color command-line  This typically happens on Suns and other systems that use shared
368  option which may be used to force Emacs to use one of a few popular  libraries.  The cause is that the site has installed a version of the
369  modes for getting colors on a tty.  For example, --color=ansi8 sets up  shared library which uses a name server--but has not installed a
370  for using the ANSI-standard escape sequences that support 8 colors.  similar version of the unshared library which Emacs uses.
371    
372  Some modes do not use colors unless you turn on the Font-lock mode.  The result is that most programs, using the shared library, work with
373  Some people have long ago set their `~/.emacs' files to turn on  the nameserver, but Emacs does not.
 Font-lock on X only, so they won't see colors on a tty.  The  
 recommended way of turning on Font-lock is by typing "M-x  
 global-font-lock-mode RET" or by customizing the variable  
 `global-font-lock-mode'.  
374    
375  * Emacs on a tty switches the cursor to large blinking block.  The fix is to install an unshared library that corresponds to what you
376    installed in the shared library, and then relink Emacs.
377    
378  This was reported to happen on some GNU/Linux systems which use  On SunOS 4.1, simply define HAVE_RES_INIT.
 ncurses version 5.0, but could be relevant for other versions as well.  
 These versions of ncurses come with a `linux' terminfo entry, where  
 the "cvvis" capability (termcap "vs") is defined as "\E[?25h\E[?8c"  
 (show cursor, change size).  This escape sequence switches on a  
 blinking hardware text-mode cursor whose size is a full character  
 cell.  This blinking cannot be stopped, since a hardware cursor  
 always blinks.  
379    
380  A work-around is to redefine the "cvvis" capability so that it  If you have already installed the name resolver in the file libresolv.a,
381  enables a *software* cursor.  The software cursor works by inverting  then you need to compile Emacs to use that library.  The easiest way to
382  the colors of the character at point, so what you see is a block  do this is to add to config.h a definition of LIBS_SYSTEM, LIBS_MACHINE
383  cursor that doesn't blink.  For this to work, you need to redefine  or LIB_STANDARD which uses -lresolv.  Watch out!  If you redefine a macro
384  the "cnorm" capability as well, so that it operates on the software  that is already in use in your configuration to supply some other libraries,
385  cursor instead of the hardware cursor.  be careful not to lose the others.
386    
387  To this end, run "infocmp linux > linux-term", edit the file  Thus, you could start by adding this to config.h:
 `linux-term' to make both the "cnorm" and "cvvis" capabilities send  
 the sequence "\E[?25h\E[?17;0;64c", and then run "tic linux-term" to  
 produce a modified terminfo entry.  
388    
389  Alternatively, if you want a blinking underscore as your Emacs cursor,  #define LIBS_SYSTEM -lresolv
 change the "cvvis" capability to send the "\E[?25h\E[?0c" command.  
390    
391  * Problems in Emacs built with LessTif.  Then if this gives you an error for redefining a macro, and you see that
392    the s- file defines LIBS_SYSTEM as -lfoo -lbar, you could change config.h
393    again to say this:
394    
395  The problems seem to depend on the version of LessTif and the Motif  #define LIBS_SYSTEM -lresolv -lfoo -lbar
 emulation for which it is set up.  
396    
397  Only the Motif 1.2 emulation seems to be stable enough in LessTif.  *** Emacs does not know your host's fully-qualified domain name.
 Lesstif 0.92-17's Motif 1.2 emulation seems to work okay on FreeBSD.  
 On GNU/Linux systems, lesstif-0.92.6 configured with "./configure  
 --enable-build-12 --enable-default-12" is reported to be the most  
 successful.  The binary GNU/Linux package  
 lesstif-devel-0.92.0-1.i386.rpm was reported to have problems with  
 menu placement.  
398    
399  On some systems, even with Motif 1.2 emulation, Emacs occasionally  You need to configure your machine with a fully qualified domain name,
400  locks up, grabbing all mouse and keyboard events.  We still don't know  either in /etc/hosts, /etc/hostname, the NIS, or wherever your system
401  what causes these problems; they are not reproducible by Emacs  calls for specifying this.
 developers.  
402    
403  * Known problems with the MS-Windows port of Emacs 21.2.  If you cannot fix the configuration, you can set the Lisp variable
404    mail-host-address to the value you want.
405    
406  Frames are not refreshed while the File or Font dialog or a pop-up menu  ** NFS and RFS
 is displayed. This also means help text for pop-up menus is not  
 displayed at all.  This is because message handling under Windows is  
 synchronous, so we cannot handle repaint (or any other) messages while  
 waiting for a system function to return the result of the dialog or  
 pop-up menu interaction.  
407    
408  Windows 95 and Windows NT up to version 4.0 do not support help text  *** Emacs says it has saved a file, but the file does not actually
409  for menus.  Help text is only available in later versions of Windows.  appear on disk.
410    
411  There are problems with display if mouse-tracking is enabled and the  This can happen on certain systems when you are using NFS, if the
412  mouse is moved off a frame, over another frame then back over the first  remote disk is full.  It is due to a bug in NFS (or certain NFS
413  frame.  A workaround is to click the left mouse button inside the frame  implementations), and there is apparently nothing Emacs can do to
414  after moving back into it.  detect the problem.  Emacs checks the failure codes of all the system
415    calls involved in writing a file, including `close'; but in the case
416    where the problem occurs, none of those system calls fails.
417    
418  Some minor flickering still persists during mouse-tracking, although  *** Editing files through RFS gives spurious "file has changed" warnings.
419  not as severely as in 21.1.  It is possible that a change in Emacs 18.37 gets around this problem,
420    but in case not, here is a description of how to fix the RFS bug that
421    causes it.
422    
423  Emacs can sometimes abort when non-ASCII text, possibly with null      There was a serious pair of bugs in the handling of the fsync() system
424  characters, is copied and pasted into a buffer.      call in the RFS server.
425    
426  An inactive cursor remains in an active window after the Windows      The first is that the fsync() call is handled as another name for the
427  Manager driven switch of the focus, until a key is pressed.      close() system call (!!).  It appears that fsync() is not used by very
428        many programs; Emacs version 18 does an fsync() before closing files
429        to make sure that the bits are on the disk.
430    
431  Windows input methods are not recognized by Emacs (as of v21.2).  Some      This is fixed by the enclosed patch to the RFS server.
 of these input methods cause the keyboard to send characters encoded  
 in the appropriate coding system (e.g., ISO 8859-1 for Latin-1  
 characters, ISO 8859-8 for Hebrew characters, etc.).  To make this  
 work, set the keyboard coding system to the appropriate value after  
 you activate the Windows input method.  For example, if you activate  
 the Hebrew input method, type "C-x RET k iso-8859-8 RET".  (Emacs  
 ought to recognize the Windows language-change event and set up the  
 appropriate keyboard encoding automatically, but it doesn't do that  
 yet.)  
432    
433  The %b specifier for format-time-string does not produce abbreviated      The second, more serious problem, is that fsync() is treated as a
434  month names with consistent widths for some locales on some versions      non-blocking system call (i.e., it's implemented as a message that
435  of Windows. This is caused by a deficiency in the underlying system      gets sent to the remote system without waiting for a reply).  Fsync is
436  library function.      a useful tool for building atomic file transactions.  Implementing it
437        as a non-blocking RPC call (when the local call blocks until the sync
438        is done) is a bad idea; unfortunately, changing it will break the RFS
439        protocol.  No fix was supplied for this problem.
440    
441  * The `configure' script doesn't find the jpeg library.      (as always, your line numbers may vary)
442    
443  There are reports that this happens on some systems because the linker      % rcsdiff -c -r1.2 serversyscall.c
444  by default only looks for shared libraries, but jpeg distribution by      RCS file: RCS/serversyscall.c,v
445  default only installs a nonshared version of the library, `libjpeg.a'.      retrieving revision 1.2
446        diff -c -r1.2 serversyscall.c
447        *** /tmp/,RCSt1003677   Wed Jan 28 15:15:02 1987
448        --- serversyscall.c     Wed Jan 28 15:14:48 1987
449        ***************
450        *** 163,169 ****
451                /*
452                 * No return sent for close or fsync!
453                 */
454        !       if (syscall == RSYS_close || syscall == RSYS_fsync)
455                        proc->p_returnval = deallocate_fd(proc, msg->m_args[0]);
456                else
457                {
458        --- 166,172 ----
459                /*
460                 * No return sent for close or fsync!
461                 */
462        !       if (syscall == RSYS_close)
463                        proc->p_returnval = deallocate_fd(proc, msg->m_args[0]);
464                else
465                {
466    
467  If this is the problem, you can configure the jpeg library with the  ** PSGML
 `--enable-shared' option and then rebuild libjpeg.  This produces a  
 shared version of libjpeg, which you need to install.  Finally, rerun  
 the Emacs configure script, which should now find the jpeg library.  
 Alternatively, modify the generated src/Makefile to link the .a file  
 explicitly, and edit src/config.h to define HAVE_JPEG.  
468    
469  * Building Emacs over NFS fails with ``Text file busy''.  *** Old versions of the PSGML package use the obsolete variables
470    `before-change-function' and `after-change-function', which are no
471    longer used by Emacs.  Please use PSGML 1.2.3 or later.
472    
473  This was reported to happen when building Emacs on a GNU/Linux system  *** PSGML conflicts with sgml-mode.
 (RedHat Linux 6.2) using a build directory automounted from Solaris  
 (SunOS 5.6) file server, but it might not be limited to that  
 configuration alone.  Presumably, the NFS server doesn't commit the  
 files' data to disk quickly enough, and the Emacs executable file is  
 left ``busy'' for several seconds after Emacs has finished dumping  
 itself.  This causes the subsequent commands which invoke the dumped  
 Emacs executable to fail with the above message.  
474    
475  In some of these cases, a time skew between the NFS server and the  PSGML package uses the same names of some variables (like keymap)
476  machine where Emacs is built is detected and reported by GNU Make  as built-in sgml-mode.el because it was created as a replacement
477  (it says that some of the files have modification time in the future).  of that package.  The conflict will be shown if you load
478  This might be a symptom of NFS-related problems.  sgml-mode.el before psgml.el.  E.g. this could happen if you edit
479    HTML page and then start to work with SGML or XML file.  html-mode
480    (from sgml-mode.el) is used for HTML file and loading of psgml.el
481    (for sgml-mode or xml-mode) will cause an error.
482    
483  If the NFS server runs on Solaris, apply the Solaris patch 105379-05  *** Versions of the PSGML package earlier than 1.0.3 (stable) or 1.1.2
484  (Sunos 5.6: /kernel/misc/nfssrv patch).  If that doesn't work, or if  (alpha) fail to parse DTD files correctly in Emacs 20.3 and later.
485  you have a different version of the OS or the NFS server, you can  Here is a patch for psgml-parse.el from PSGML 1.0.1 and, probably,
486  force the NFS server to use 1KB blocks, which was reported to fix the  earlier versions.
 problem albeit at a price of slowing down file I/O.  You can force 1KB  
 blocks by specifying the "-o  rsize=1024,wsize=1024" options to the  
 `mount' command, or by adding ",rsize=1024,wsize=1024" to the mount  
 options in the appropriate system configuration file, such as  
 `/etc/auto.home'.  
487    
488  Alternatively, when Make fails due to this problem, you could wait for  --- psgml-parse.el      1998/08/21 19:18:18     1.1
489  a few seconds and then invoke Make again.  In one particular case,  +++ psgml-parse.el      1998/08/21 19:20:00
490  waiting for 10 or more seconds between the two Make invocations seemed  @@ -2383,7 +2383,7 @@ (defun sgml-push-to-entity (entity &opti
491  to work around the problem.         (setq sgml-buffer-parse-state nil))
492         (cond
493          ((stringp entity)                 ; a file name
494    -      (save-excursion (insert-file-contents entity))
495    +      (insert-file-contents entity)
496           (setq default-directory (file-name-directory entity)))
497          ((consp (sgml-entity-text entity)) ; external id?
498           (let* ((extid (sgml-entity-text entity))
499    
500  Similar problems can happen if your machine NFS-mounts a directory  ** AUCTeX
 onto itself.  Suppose the Emacs sources live in `/usr/local/src' and  
 you are working on the host called `marvin'.  Then an entry in the  
 `/etc/fstab' file like the following is asking for trouble:  
501    
502      marvin:/usr/local/src /usr/local/src ...options.omitted...  You should not be using a version older than 11.52 if you can avoid
503    it.
504    
505  The solution is to remove this line from `etc/fstab'.  *** Emacs 21 freezes when visiting a TeX file with AUCTeX installed.
506    
507  * Emacs binary is not in executable format, and cannot be run.  Emacs 21 needs version 10 or later of AUCTeX; upgrading should solve
508    these problems.
509    
510  This was reported to happen when Emacs is built in a directory mounted  *** No colors in AUCTeX with Emacs 21.
 via NFS, for some combinations of NFS client and NFS server.  
 Usually, the file `emacs' produced in these cases is full of  
 binary null characters, and the `file' utility says:  
511    
512      emacs: ASCII text, with no line terminators  Upgrade to AUC TeX version 10 or later, and make sure it is
513    byte-compiled with Emacs 21.
514    
515  We don't know what exactly causes this failure.  A work-around is to  ** Miscellaneous problems
 build Emacs in a directory on a local disk.  
516    
517  * Accented ISO-8859-1 characters are displayed as | or _.  *** Self-documentation messages are garbled.
518    
519  Try other font set sizes (S-mouse-1).  If the problem persists with  This means that the file `etc/DOC-...' doesn't properly correspond
520  other sizes as well, your text is corrupted, probably through software  with the Emacs executable.  Redumping Emacs and then installing the
521  that is not 8-bit clean.  If the problem goes away with another font  corresponding pair of files should fix the problem.
 size, it's probably because some fonts pretend to be ISO-8859-1 fonts  
 when they are really ASCII fonts. In particular the schumacher-clean  
 fonts have this bug in some versions of X.  
522    
523  To see what glyphs are included in a font, use `xfd', like this:  *** Programs running under terminal emulator do not recognize `emacs'
524    terminal type.
525    
526    xfd -fn -schumacher-clean-medium-r-normal--12-120-75-75-c-60-iso8859-1  The cause of this is a shell startup file that sets the TERMCAP
527    environment variable.  The terminal emulator uses that variable to
528    provide the information on the special terminal type that Emacs
529    emulates.
530    
531  If this shows only ASCII glyphs, the font is indeed the source of the  Rewrite your shell startup file so that it does not change TERMCAP
532  problem.  in such a case.  You could use the following conditional which sets
533    it only if it is undefined.
534    
535  The solution is to remove the corresponding lines from the appropriate      if ( ! ${?TERMCAP} ) setenv TERMCAP ~/my-termcap-file
 `fonts.alias' file, then run `mkfontdir' in that directory, and then run  
 `xset fp rehash'.  
536    
537  * Large file support is disabled on HP-UX.  See the comments in  Or you could set TERMCAP only when you set TERM--which should not
538  src/s/hpux10.h.  happen in a non-login shell.
539    
540  * Crashes when displaying GIF images in Emacs built with version  *** In Shell mode, you get a ^M at the end of every line.
 libungif-4.1.0 are resolved by using version libungif-4.1.0b1.  
 Configure checks for the correct version, but this problem could occur  
 if a binary built against a shared libungif is run on a system with an  
 older version.  
541    
542  * Font Lock displays portions of the buffer in incorrect faces.  This happens to people who use tcsh, because it is trying to be too
543    smart.  It sees that the Shell uses terminal type `unknown' and turns
544    on the flag to output ^M at the end of each line.  You can fix the
545    problem by adding this to your .cshrc file:
546    
547  By far the most frequent cause of this is a parenthesis `(' or a brace      if ($?EMACS) then
548  `{' in column zero.  Font Lock assumes that such a paren is outside of          if ($EMACS == "t") then
549  any comment or string.  This is of course not true in general, but the              unset edit
550  vast majority of well-formatted program source files don't have such              stty  -icrnl -onlcr -echo susp ^Z
551  parens, and therefore this assumption is used to allow optimizations          endif
552  in Font Lock's syntactical analysis.  These optimizations avoid some      endif
 pathological cases where jit-lock, the Just-in-Time fontification  
 introduced with Emacs 21.1, could significantly slow down scrolling  
 through the buffer, especially scrolling backwards, and also jumping  
 to the end of a very large buffer.  
553    
554  Beginning with version 21.4, a parenthesis or a brace in column zero  *** Emacs startup on GNU/Linux systems (and possibly other systems) is slow.
 is highlighted in bold-red face if it is inside a string or a comment,  
 to indicate that it could interfere with Font Lock (and also with  
 indentation) and should be moved or escaped with a backslash.  
555    
556  If you don't use large buffers, or have a very fast machine which  This can happen if the system is misconfigured and Emacs can't get the
557  makes the delays insignificant, you can avoid the incorrect  full qualified domain name, FQDN.  You should have your FQDN in the
558  fontification by setting the variable  /etc/hosts file, something like this:
 `font-lock-beginning-of-syntax-function' to a nil value.  (This must  
 be done _after_ turning on Font Lock.)  
559    
560  Another alternative is to avoid a paren in column zero.  For example,  127.0.0.1       localhost
561  in a Lisp string you could precede the paren with a backslash.  129.187.137.82  nuc04.t30.physik.tu-muenchen.de nuc04
562    
563  * When running on KDE, colors or fonts are not as specified for Emacs,  The way to set this up may vary on non-GNU systems.
 or messed up.  
564    
565  For example, you could see background you set for Emacs only in the  *** Attempting to visit remote files via ange-ftp fails.
 empty portions of the Emacs display, while characters have some other  
 background.  
566    
567  This happens because KDE's defaults apply its color and font  If the error message is "ange-ftp-file-modtime: Specified time is not
568  definitions even to applications that weren't compiled for KDE.  The  representable", then this could happen when `lukemftp' is used as the
569  solution is to uncheck the "Apply fonts and colors to non-KDE apps"  ftp client.  This was reported to happen on Debian GNU/Linux, kernel
570  option in Preferences->Look&Feel->Style (KDE 2).  In KDE 3, this option  version 2.4.3, with `lukemftp' 1.5-5, but might happen on other
571  is in the "Colors" section, rather than "Style".  systems as well.  To avoid this problem, switch to using the standard
572    ftp client.  On a Debian system, type
573    
574  Alternatively, if you do want the KDE defaults to apply to other    update-alternatives --config ftp
 applications, but not to Emacs, you could modify the file `Emacs.ad'  
 (should be in the `/usr/share/apps/kdisplay/app-defaults/' directory)  
 so that it doesn't set the default background and foreground only for  
 Emacs.  For example, make sure the following resources are either not  
 present or commented out:  
575    
576     Emacs.default.attributeForeground  and then choose /usr/bin/netkit-ftp.
    Emacs.default.attributeBackground  
    Emacs*Foreground  
    Emacs*Background  
577    
578  * Interrupting Cygwin port of Bash from Emacs doesn't work.  *** JPEG images aren't displayed.
579    
580  Cygwin 1.x builds of the ported Bash cannot be interrupted from the  This has been reported when Emacs is built with jpeg-6a library.
581  MS-Windows version of Emacs.  This is due to some change in the Bash  Upgrading to jpeg-6b solves the problem.  Configure checks for the
582  port or in the Cygwin library which apparently make Bash ignore the  correct version, but this problem could occur if a binary built
583  keyboard interrupt event sent by Emacs to Bash.  (Older Cygwin ports  against a shared libjpeg is run on a system with an older version.
 of Bash, up to b20.1, did receive SIGINT from Emacs.)  
584    
585  * Dired is very slow.  *** Dired is very slow.
586    
587  This could happen if invocation of the `df' program takes a long  This could happen if invocation of the `df' program takes a long
588  time.  Possible reasons for this include:  time.  Possible reasons for this include:
# Line 768  To work around the problem, you could ei Line 599  To work around the problem, you could ei
599  invoking `df'; (b) use `df' from the GNU Fileutils package; or  invoking `df'; (b) use `df' from the GNU Fileutils package; or
600  (c) use CVS, which is Free Software, instead of ClearCase.  (c) use CVS, which is Free Software, instead of ClearCase.
601    
602  * Accessing remote files with ange-ftp hangs the MS-Windows version of Emacs.  *** Versions of the W3 package released before Emacs 21.1 don't run
   
 If the FTP client is the Cygwin port of GNU `ftp', this appears to be  
 due to some bug in the Cygwin DLL or some incompatibility between it  
 and the implementation of asynchronous subprocesses in the Windows  
 port of Emacs.  Specifically, some parts of the FTP server responses  
 are not flushed out, apparently due to buffering issues, which  
 confuses ange-ftp.  
   
 The solution is to downgrade to an older version of the Cygwin DLL  
 (version 1.3.2 was reported to solve the problem), or use the stock  
 Windows FTP client, usually found in the `C:\WINDOWS' or 'C:\WINNT'  
 directory.  To force ange-ftp use the stock Windows client, set the  
 variable `ange-ftp-ftp-program-name' to the absolute file name of the  
 client's executable.  For example:  
   
  (setq ange-ftp-ftp-program-name "c:/windows/ftp.exe")  
   
 If you want to stick with the Cygwin FTP client, you can work around  
 this problem by putting this in your `.emacs' file:  
   
  (setq ange-ftp-ftp-program-args '("-i" "-n" "-g" "-v" "--prompt" "")  
   
 * Versions of the W3 package released before Emacs 21.1 don't run  
603  under Emacs 21.  This fixed in W3 version 4.0pre.47.  under Emacs 21.  This fixed in W3 version 4.0pre.47.
604    
605  * On AIX, if linking fails because libXbsd isn't found, check if you  *** The LDAP support rely on ldapsearch program from OpenLDAP version 2.
 are compiling with the system's `cc' and CFLAGS containing `-O5'.  If  
 so, you have hit a compiler bug.  Please make sure to re-configure  
 Emacs so that it isn't compiled with `-O5'.  
   
 * Compiling on AIX 4.3.x or 4.4 fails.  
   
 This could happen if you use /bin/c89 as your compiler, instead of  
 the default `cc'.  /bin/c89 treats certain warnings, such as benign  
 redefinitions of macros, as errors, and fails the build.  A solution  
 is to use the default compiler `cc'.  
   
 * Old versions of the PSGML package use the obsolete variables  
 `before-change-function' and `after-change-function', which are no  
 longer used by Emacs.  Please use PSGML 1.2.3 or later.  
   
 * PSGML conflicts with sgml-mode.  
   
 PSGML package uses the same names of some variables (like keymap)  
 as built-in sgml-mode.el because it was created as a replacement  
 of that package.  The conflict will be shown if you load  
 sgml-mode.el before psgml.el.  E.g. this could happen if you edit  
 HTML page and then start to work with SGML or XML file.  html-mode  
 (from sgml-mode.el) is used for HTML file and loading of psgml.el  
 (for sgml-mode or xml-mode) will cause an error.  
   
 * The LDAP support rely on ldapsearch program from OpenLDAP version 2.  
606    
607  It can fail to work with ldapsearch program from OpenLDAP version 1.  It can fail to work with ldapsearch program from OpenLDAP version 1.
608  Version 1 of OpenLDAP is now deprecated.  If you are still using it,  Version 1 of OpenLDAP is now deprecated.  If you are still using it,
609  please upgrade to version 2.  As a temporary workaround, remove  please upgrade to version 2.  As a temporary workaround, remove
610  argument "-x" from the variable `ldap-ldapsearch-args'.  argument "-x" from the variable `ldap-ldapsearch-args'.
611    
612  * The `oc-unicode' package doesn't work with Emacs 21.  *** ps-print commands fail to find prologue files ps-prin*.ps.
   
 This package tries to define more private charsets than there are free  
 slots now.  The current built-in Unicode support is actually more  
 flexible.  (Use option `utf-translate-cjk-mode' if you need CJK  
 support.)  Files encoded as emacs-mule using oc-unicode aren't  
 generally read correctly by Emacs 21.  
   
 * Using epop3.el package causes Emacs to signal an error.  
   
 The error message might be something like this:  
   
   "Lisp nesting exceeds max-lisp-eval-depth"  
   
 This happens because epop3 redefines the function gethash, which is a  
 built-in primitive beginning with Emacs 21.1.  We don't have a patch  
 for epop3 that fixes this, but perhaps a newer version of epop3  
 corrects that.  
   
 * ps-print commands fail to find prologue files ps-prin*.ps.  
613    
614  This can happen if you use an old version of X-Symbol package: it  This can happen if you use an old version of X-Symbol package: it
615  defines compatibility functions which trick ps-print into thinking it  defines compatibility functions which trick ps-print into thinking it
# Line 854  runs in XEmacs, and look for the prologu Line 617  runs in XEmacs, and look for the prologu
617    
618  The solution is to upgrade X-Symbol to a later version.  The solution is to upgrade X-Symbol to a later version.
619    
620  * lpr commands don't work on MS-Windows with some cheap printers.  *** On systems with shared libraries you might encounter run-time errors
   
 This problem may also strike other platforms, but the solution is  
 likely to be a global one, and not Emacs specific.  
   
 Many cheap inkjet, and even some cheap laser printers, do not  
 print plain text anymore, they will only print through graphical  
 printer drivers. A workaround on MS-Windows is to use Windows' basic  
 built in editor to print (this is possibly the only useful purpose it  
 has):  
   
 (setq printer-name "")         ;; notepad takes the default  
 (setq lpr-command "notepad")   ;; notepad  
 (setq lpr-switches nil)        ;; not needed  
 (setq lpr-printer-switch "/P") ;; run notepad as batch printer  
   
 * On systems with shared libraries you might encounter run-time errors  
621  from the dynamic linker telling you that it is unable to find some  from the dynamic linker telling you that it is unable to find some
622  shared libraries, for instance those for Xaw3d or image support.  shared libraries, for instance those for Xaw3d or image support.
623  These errors mean Emacs has been linked with a library whose shared  These errors mean Emacs has been linked with a library whose shared
# Line 909  to work around the problem. Line 656  to work around the problem.
656    
657  Please refer to the documentation of your dynamic linker for details.  Please refer to the documentation of your dynamic linker for details.
658    
659  * On Solaris 2.7, building Emacs with WorkShop Compilers 5.0 98/12/15  *** You request inverse video, and the first Emacs frame is in inverse
660  C 5.0 failed, apparently with non-default CFLAGS, most probably due to  video, but later frames are not in inverse video.
 compiler bugs.  Using Sun Solaris 2.7 Sun WorkShop 6 update 1 C  
 release was reported to work without problems.  It worked OK on  
 another system with Solaris 8 using apparently the same 5.0 compiler  
 and the default CFLAGS.  
   
 * Compiling syntax.c with the OPENSTEP 4.2 compiler gcc 2.7.2.1 fails.  
   
 The compiler was reported to crash while compiling syntax.c with the  
 following message:  
   
    cc: Internal compiler error: program cc1obj got fatal signal 11  
   
 To work around this, replace the macros UPDATE_SYNTAX_TABLE_FORWARD,  
 INC_BOTH, and INC_FROM with functions.  To this end, first define 3  
 functions, one each for every macro.  Here's an example:  
   
     static int update_syntax_table_forward(int from)  
     {  
         return(UPDATE_SYNTAX_TABLE_FORWARD(from));  
     }/*update_syntax_table_forward*/  
661    
662  Then replace all references to UPDATE_SYNTAX_TABLE_FORWARD in syntax.c  This can happen if you have an old version of the custom library in
663  with a call to the function update_syntax_table_forward.  your search path for Lisp packages.  Use M-x list-load-path-shadows to
664    check whether this is true.  If it is, delete the old custom library.
665    
666  * Emacs fails to start, complaining about missing fonts.  *** When you run Ispell from Emacs, it reports a "misalignment" error.
667    
668  A typical error message might be something like  This can happen if you compiled the Ispell program to use ASCII
669    characters only and then try to use it from Emacs with non-ASCII
670    characters, like Latin-1.  The solution is to recompile Ispell with
671    support for 8-bit characters.
672    
673    No fonts match `-*-fixed-medium-r-*--6-*-*-*-*-*-iso8859-1'  To see whether your Ispell program supports 8-bit characters, type
674    this at your shell's prompt:
675    
676  This happens because some X resource specifies a bad font family for       ispell -vv
 Emacs to use.  The possible places where this specification might be  
 are:  
677    
678    - in your ~/.Xdefaults file  and look in the output for the string "NO8BIT".  If Ispell says
679    "!NO8BIT (8BIT)", your speller supports 8-bit characters; otherwise it
680    does not.
681    
682    - client-side X resource file, such as  ~/Emacs or  To rebuild Ispell with 8-bit character support, edit the local.h file
683      /usr/X11R6/lib/app-defaults/Emacs or  in the Ispell distribution and make sure it does _not_ define NO8BIT.
684      /usr/X11R6/lib/X11/app-defaults/Emacs  Then rebuild the speller.
685    
686  One of these files might have bad or malformed specification of a  Another possible cause for "misalignment" error messages is that the
687  fontset that Emacs should use.  To fix the problem, you need to find  version of Ispell installed on your machine is old.  Upgrade.
 the problematic line(s) and correct them.  
688    
689  * Emacs 20 and later fails to load Lisp files at startup.  Yet another possibility is that you are trying to spell-check a word
690    in a language that doesn't fit the dictionary you choose for use by
691    Ispell.  (Ispell can only spell-check one language at a time, because
692    it uses a single dictionary.)  Make sure that the text you are
693    spelling and the dictionary used by Ispell conform to each other.
694    
695  The typical error message might be like this:  If your spell-checking program is Aspell, it has been reported that if
696    you have a personal configuration file (normally ~/.aspell.conf), it
697    can cause this error.  Remove that file, execute `ispell-kill-ispell'
698    in Emacs, and then try spell-checking again.
699    
700    "Cannot open load file: fontset"  * Runtime problems related to font handling
701    
702  This could happen if you compress the file lisp/subdirs.el.  That file  ** Under X11, some characters appear as hollow boxes.
 tells Emacs what are the directories where it should look for Lisp  
 files.  Emacs cannot work with subdirs.el compressed, since the  
 Auto-compress mode it needs for this will not be loaded until later,  
 when your .emacs file is processed.  (The package `fontset.el' is  
 required to set up fonts used to display text on window systems, and  
 it's loaded very early in the startup procedure.)  
703    
704  Similarly, any other .el file for which there's no corresponding .elc  Each X11 font covers just a fraction of the characters that Emacs
705  file could fail to load if it is compressed.  supports.  To display the whole range of Emacs characters requires
706    many different fonts, collected into a fontset.
707    
708  The solution is to uncompress all .el files which don't have a .elc  If some of the fonts called for in your fontset do not exist on your X
709  file.  server, then the characters that have no font appear as hollow boxes.
710    You can remedy the problem by installing additional fonts.
711    
712  Another possible reason for such failures is stale *.elc files  The intlfonts distribution includes a full spectrum of fonts that can
713  lurking somewhere on your load-path.  The following command will  display all the characters Emacs supports.
 print any duplicate Lisp files that are present in load-path:  
714    
715      emacs -q -batch -f list-load-path-shadows  Another cause of this for specific characters is fonts which have a
716    missing glyph and no default character.  This is known to occur for
717    character number 160 (no-break space) in some fonts, such as Lucida
718    but Emacs sets the display table for the unibyte and Latin-1 version
719    of this character to display a space.
720    
721  If this command prints any file names, some of these files are stale,  ** Under X11, some characters appear improperly aligned in their lines.
 and should be deleted or their directories removed from your  
 load-path.  
722    
723  * Emacs prints an error at startup after upgrading from an earlier version.  You may have bad X11 fonts; try installing the intlfonts distribution.
724    
725  An example of such an error is:  ** Certain fonts make each line take one pixel more than it "should".
726    
727    x-complement-fontset-spec: "Wrong type argument: stringp, nil"  This is because these fonts contain characters a little taller
728    than the font's nominal height.  Emacs needs to make sure that
729    lines do not overlap.
730    
731  This can be another symptom of stale *.elc files in your load-path.  ** Loading fonts is very slow.
 The following command will print any duplicate Lisp files that are  
 present in load-path:  
732    
733      emacs -q -batch -f list-load-path-shadows  You might be getting scalable fonts instead of precomputed bitmaps.
734    Known scalable font directories are "Type1" and "Speedo".  A font
735    directory contains scalable fonts if it contains the file
736    "fonts.scale".
737    
738  If this command prints any file names, some of these files are stale,  If this is so, re-order your X windows font path to put the scalable
739  and should be deleted or their directories removed from your  font directories last.  See the documentation of `xset' for details.
 load-path.  
740    
741  * Attempting to visit remote files via ange-ftp fails.  With some X servers, it may be necessary to take the scalable font
742    directories out of your path entirely, at least for Emacs 19.26.
743    Changes in the future may make this unnecessary.
744    
745  If the error message is "ange-ftp-file-modtime: Specified time is not  ** Font Lock displays portions of the buffer in incorrect faces.
 representable", then this could happen when `lukemftp' is used as the  
 ftp client.  This was reported to happen on Debian GNU/Linux, kernel  
 version 2.4.3, with `lukemftp' 1.5-5, but might happen on other  
 systems as well.  To avoid this problem, switch to using the standard  
 ftp client.  On a Debian system, type  
746    
747    update-alternatives --config ftp  By far the most frequent cause of this is a parenthesis `(' or a brace
748    `{' in column zero.  Font Lock assumes that such a paren is outside of
749    any comment or string.  This is of course not true in general, but the
750    vast majority of well-formatted program source files don't have such
751    parens, and therefore this assumption is used to allow optimizations
752    in Font Lock's syntactical analysis.  These optimizations avoid some
753    pathological cases where jit-lock, the Just-in-Time fontification
754    introduced with Emacs 21.1, could significantly slow down scrolling
755    through the buffer, especially scrolling backwards, and also jumping
756    to the end of a very large buffer.
757    
758  and then choose /usr/bin/netkit-ftp.  Beginning with version 21.4, a parenthesis or a brace in column zero
759    is highlighted in bold-red face if it is inside a string or a comment,
760    to indicate that it could interfere with Font Lock (and also with
761    indentation) and should be moved or escaped with a backslash.
762    
763  * Antivirus software interacts badly with the MS-Windows version of Emacs.  If you don't use large buffers, or have a very fast machine which
764    makes the delays insignificant, you can avoid the incorrect
765    fontification by setting the variable
766    `font-lock-beginning-of-syntax-function' to a nil value.  (This must
767    be done _after_ turning on Font Lock.)
768    
769  The usual manifestation of these problems is that subprocesses don't  Another alternative is to avoid a paren in column zero.  For example,
770  work or even wedge the entire system.  In particular, "M-x shell RET"  in a Lisp string you could precede the paren with a backslash.
 was reported to fail to work.  But other commands also sometimes don't  
 work when an antivirus package is installed.  
771    
772  The solution is to switch the antivirus software to a less aggressive  ** With certain fonts, when the cursor appears on a character, the
773  mode (e.g., disable the ``auto-protect'' feature), or even uninstall  character doesn't appear--you get a solid box instead.
 or disable it entirely.  
774    
775  * On MS-Windows 95/98/ME, subprocesses do not terminate properly.  One user on a Linux-based GNU system reported that this problem went
776    away with installation of a new X server.  The failing server was
777    XFree86 3.1.1.  XFree86 3.1.2 works.
778    
779  This is a limitation of the Operating System, and can cause problems  ** Characters are displayed as empty boxes or with wrong font under X.
 when shutting down Windows. Ensure that all subprocesses are exited  
 cleanly before exiting Emacs. For more details, see the FAQ at  
 http://www.gnu.org/software/emacs/windows/.  
780    
781  * MS-Windows 95/98/ME crashes when Emacs invokes non-existent programs.  This can occur when two different versions of FontConfig are used.
782    For example, XFree86 4.3.0 has one version and Gnome usually comes
783    with a newer version.  Emacs compiled with --with-gtk will then use
784    the newer version.  In most cases the problem can be temporarily
785    fixed by stopping the application that has the error (it can be
786    Emacs or any other application), removing ~/.fonts.cache-1,
787    and then start the application again.
788    If removing ~/.fonts.cache-1 and restarting doesn't help, the
789    application with problem must be recompiled with the same version
790    of FontConfig as the rest of the system uses.  For KDE, it is
791    sufficient to recompile Qt.
792    
793  When a program you are trying to run is not found on the PATH,  ** Emacs pauses for several seconds when changing the default font.
 Windows might respond by crashing or locking up your system.  In  
 particular, this has been reported when trying to compile a Java  
 program in JDEE when javac.exe is installed, but not on the system  
 PATH.  
794    
795  * Pressing the mouse button on MS-Windows does not give a mouse-2 event.  This has been reported for fvwm 2.2.5 and the window manager of KDE
796    2.1.  The reason for the pause is Xt waiting for a ConfigureNotify
797    event from the window manager, which the window manager doesn't send.
798    Xt stops waiting after a default timeout of usually 5 seconds.
799    
800  This is usually a problem with the mouse driver. Because most Windows  A workaround for this is to add something like
 programs do not do anything useful with the middle mouse button, many  
 mouse drivers allow you to define the wheel press to do something  
 different. Some drivers do not even have the option to generate a  
 middle button press. In such cases, setting the wheel press to  
 "scroll" sometimes works if you press the button twice. Trying a  
 generic mouse driver might help.  
801    
802  * Scrolling the mouse wheel on MS-Windows always scrolls the top window.  emacs.waitForWM: false
803    
804  This is another common problem with mouse drivers. Instead of  to your X resources.  Alternatively, add `(wait-for-wm . nil)' to a
805  generating scroll events, some mouse drivers try to fake scroll bar  frame's parameter list, like this:
 movement. But they are not intelligent enough to handle multiple  
 scroll bars within a frame. Trying a generic mouse driver might help.  
806    
807  * Mail sent through Microsoft Exchange in some encodings appears to be     (modify-frame-parameters nil '((wait-for-wm . nil)))
 mangled and is not seen correctly in Rmail or Gnus.  We don't know  
 exactly what happens, but it isn't an Emacs problem in cases we've  
 seen.  
808    
809  * After upgrading to a newer version of Emacs, the Meta key stops working.  (this should go into your `.emacs' file).
810    
811  This was reported to happen on a GNU/Linux system distributed by  ** Underlines appear at the wrong position.
 Mandrake.  The reason is that the previous version of Emacs was  
 modified by Mandrake to make the Alt key act as the Meta key, on a  
 keyboard where the Windows key is the one which produces the Meta  
 modifier.  A user who started using a newer version of Emacs, which  
 was not hacked by Mandrake, expected the Alt key to continue to act as  
 Meta, and was astonished when that didn't happen.  
812    
813  The solution is to find out what key on your keyboard produces the Meta  This is caused by fonts having a wrong UNDERLINE_POSITION property.
814  modifier, and use that key instead.  Try all of the keys to the left  Examples are the font 7x13 on XFree prior to version 4.1, or the jmk
815  and to the right of the space bar, together with the `x' key, and see  neep font from the Debian xfonts-jmk package.  To circumvent this
816  which combination produces "M-x" in the echo area.  You can also use  problem, set x-use-underline-position-properties to nil in your
817  the `xmodmap' utility to show all the keys which produce a Meta  `.emacs'.
 modifier:  
818    
819           xmodmap -pk | egrep -i "meta|alt"  To see what is the value of UNDERLINE_POSITION defined by the font,
820    type `xlsfonts -lll FONT' and look at the font's UNDERLINE_POSITION
821    property.
822    
823  A more convenient way of finding out which keys produce a Meta modifier  ** When using Exceed, fonts sometimes appear too tall.
 is to use the `xkbprint' utility, if it's available on your system:  
824    
825           xkbprint 0:0 /tmp/k.ps  When the display is set to an Exceed X-server and fonts are specified
826    (either explicitly with the -fn option or implicitly with X resources)
827    then the fonts may appear "too tall".  The actual character sizes are
828    correct but there is too much vertical spacing between rows,  which
829    gives the appearance of "double spacing".
830    
831  This produces a PostScript file `/tmp/k.ps' with a picture of your  To prevent this, turn off the Exceed's "automatic font substitution"
832  keyboard; printing that file on a PostScript printer will show what  feature (in the font part of the configuration window).
 keys can serve as Meta.  
833    
834  The `xkeycaps' also shows a visual representation of the current  * Internationalization problems
 keyboard settings.  It also allows to modify them.  
835    
836  * On OSF/Dec Unix/Tru64/<whatever it is this year> under X locally or  ** Characters from the mule-unicode charsets aren't displayed under X.
 remotely, M-SPC acts as a `compose' key with strange results.  See  
 keyboard(5).  
837    
838  Changing Alt_L to Meta_L fixes it:  XFree86 4 contains many fonts in iso10646-1 encoding which have
839  % xmodmap -e 'keysym Alt_L = Meta_L Alt_L'  minimal character repertoires (whereas the encoding part of the font
840  % xmodmap -e 'keysym Alt_R = Meta_R Alt_R'  name is meant to be a reasonable indication of the repertoire
841    according to the XLFD spec).  Emacs may choose one of these to display
842    characters from the mule-unicode charsets and then typically won't be
843    able to find the glyphs to display many characters.  (Check with C-u
844    C-x = .)  To avoid this, you may need to use a fontset which sets the
845    font for the mule-unicode sets explicitly.  E.g. to use GNU unifont,
846    include in the fontset spec:
847    
848  * Error "conflicting types for `initstate'" compiling with GCC on Irix 6.  mule-unicode-2500-33ff:-gnu-unifont-*-iso10646-1,\
849    mule-unicode-e000-ffff:-gnu-unifont-*-iso10646-1,\
850    mule-unicode-0100-24ff:-gnu-unifont-*-iso10646-1
851    
852  Install GCC 2.95 or a newer version, and this problem should go away.  ** The UTF-8/16/7 coding systems don't encode CJK (Far Eastern) characters.
 It is possible that this problem results from upgrading the operating  
 system without reinstalling GCC; so you could also try reinstalling  
 the same version of GCC, and telling us whether that fixes the problem.  
853    
854  * Emacs dumps core on Solaris in function IMCheckWindow.  Emacs by default only supports the parts of the Unicode BMP whose code
855    points are in the ranges 0000-33ff and e000-ffff.  This excludes: most
856    of CJK, Yi and Hangul, as well as everything outside the BMP.
857    
858  This was reported to happen when Emacs runs with more than one frame,  If you read UTF-8 data with code points outside these ranges, the
859  and one of them is closed, either with "C-x 5 0" or from the window  characters appear in the buffer as raw bytes of the original UTF-8
860  manager.  (composed into a single quasi-character) and they will be written back
861    correctly as UTF-8, assuming you don't break the composed sequences.
862    If you read such characters from UTF-16 or UTF-7 data, they are
863    substituted with the Unicode `replacement character', and you lose
864    information.
865    
866  This bug was reported to Sun as  To edit such UTF data, turn on Utf-Translate-Cjk mode, which makes
867    many common CJK characters available for encoding and decoding and can
868    be extended by updating the tables it uses.  This also allows you to
869    save as UTF buffers containing characters decoded by the chinese-,
870    japanese- and korean- coding systems, e.g. cut and pasted from
871    elsewhere.
872    
873      Gtk apps dump core in ximlocal.so.2:IMCheckIMWindow()  ** Mule-UCS loads very slowly.
     Bug Reports: 4463537  
874    
875  Installing Solaris 8 patch 108773-12 for Sparc and 108774-12 for x86  Changes to Emacs internals interact badly with Mule-UCS's `un-define'
876  reportedly fixes the bug, which appears to be inside the shared  library, which is the usual interface to Mule-UCS.  Apply the
877  library xiiimp.so.  following patch to Mule-UCS 0.84 and rebuild it.  That will help,
878    though loading will still be slower than in Emacs 20.  (Some
879    distributions, such as Debian, may already have applied such a patch.)
880    
881  Alternatively, you can configure Emacs with `--with-xim=no' to prevent  --- lisp/un-define.el   6 Mar 2001 22:41:38 -0000       1.30
882  the core dump, but will loose X input method support, of course.  (You  +++ lisp/un-define.el   19 Apr 2002 18:34:26 -0000
883  can use Emacs's own input methods instead, if you install Leim.)  @@ -610,13 +624,21 @@ by calling post-read-conversion and pre-
884    
885  * On Solaris 7, Emacs gets a segmentation fault when starting up using X.    (mapcar
886       (lambda (x)
887    -    (mapcar
888    -     (lambda (y)
889    -       (mucs-define-coding-system
890    -       (nth 0 y) (nth 1 y) (nth 2 y)
891    -       (nth 3 y) (nth 4 y) (nth 5 y) (nth 6 y))
892    -       (coding-system-put (car y) 'alias-coding-systems (list (car x))))
893    -     (cdr x)))
894    +    (if (fboundp 'register-char-codings)
895    +       ;; Mule 5, where we don't need the eol-type specified and
896    +       ;; register-char-codings may be very slow for these coding
897    +       ;; system definitions.
898    +       (let ((y (cadr x)))
899    +         (mucs-define-coding-system
900    +          (car x) (nth 1 y) (nth 2 y)
901    +          (nth 3 y) (nth 4 y) (nth 5 y)))
902    +      (mapcar
903    +       (lambda (y)
904    +        (mucs-define-coding-system
905    +         (nth 0 y) (nth 1 y) (nth 2 y)
906    +         (nth 3 y) (nth 4 y) (nth 5 y) (nth 6 y))
907    +        (coding-system-put (car y) 'alias-coding-systems (list (car x)))))
908    +      (cdr x)))
909       `((utf-8
910          (utf-8-unix
911           ?u "UTF-8 coding system"
912    
913  This results from Sun patch 107058-01 (SunOS 5.7: Patch for  Note that Emacs has native support for Unicode, roughly equivalent to
914  assembler) if you use GCC version 2.7 or later.  Mule-UCS's, so you may not need it.
 To work around it, either install patch 106950-03 or later,  
 or uninstall patch 107058-01, or install the GNU Binutils.  
 Then recompile Emacs, and it should work.  
915    
916  * With X11R6.4, public-patch-3, Emacs crashes at startup.  ** Accented ISO-8859-1 characters are displayed as | or _.
917    
918  Reportedly this patch in X fixes the problem.  Try other font set sizes (S-mouse-1).  If the problem persists with
919    other sizes as well, your text is corrupted, probably through software
920    that is not 8-bit clean.  If the problem goes away with another font
921    size, it's probably because some fonts pretend to be ISO-8859-1 fonts
922    when they are really ASCII fonts. In particular the schumacher-clean
923    fonts have this bug in some versions of X.
924    
925      --- xc/lib/X11/imInt.c~     Wed Jun 30 13:31:56 1999  To see what glyphs are included in a font, use `xfd', like this:
     +++ xc/lib/X11/imInt.c      Thu Jul  1 15:10:27 1999  
     @@ -1,4 +1,4 @@  
     -/* $TOG: imInt.c /main/5 1998/05/30 21:11:16 kaleb $ */  
     +/* $TOG: imInt.c /main/5 1998/05/30 21:11:16 kaleb $  */  
      /******************************************************************  
926    
927                  Copyright 1992, 1993, 1994 by FUJITSU LIMITED    xfd -fn -schumacher-clean-medium-r-normal--12-120-75-75-c-60-iso8859-1
     @@ -166,8 +166,8 @@  
      _XimMakeImName(lcd)  
          XLCd      lcd;  
      {  
     -    char* begin;  
     -    char* end;  
     +    char* begin = NULL;  
     +    char* end = NULL;  
          char* ret;  
          int    i = 0;  
          char* ximmodifier = XIMMODIFIER;  
     @@ -182,7 +182,11 @@  
          }  
          ret = Xmalloc(end - begin + 2);  
          if (ret != NULL) {  
     -           (void)strncpy(ret, begin, end - begin + 1);  
     +   if (begin != NULL) {  
     +             (void)strncpy(ret, begin, end - begin + 1);  
     +        } else {  
     +     ret[0] = '\0';  
     +   }  
             ret[end - begin + 1] = '\0';  
          }  
          return ret;  
928    
929    If this shows only ASCII glyphs, the font is indeed the source of the
930    problem.
931    
932  * Emacs crashes on Irix 6.5 on the SGI R10K, when compiled with GCC.  The solution is to remove the corresponding lines from the appropriate
933    `fonts.alias' file, then run `mkfontdir' in that directory, and then run
934    `xset fp rehash'.
935    
936  This seems to be fixed in GCC 2.95.  ** The `oc-unicode' package doesn't work with Emacs 21.
937    
938  * Emacs crashes in utmpname on Irix 5.3.  This package tries to define more private charsets than there are free
939    slots now.  The current built-in Unicode support is actually more
940    flexible.  (Use option `utf-translate-cjk-mode' if you need CJK
941    support.)  Files encoded as emacs-mule using oc-unicode aren't
942    generally read correctly by Emacs 21.
943    
944  This problem is fixed in Patch 3175 for Irix 5.3.  ** After a while, Emacs slips into unibyte mode.
 It is also fixed in Irix versions 6.2 and up.  
945    
946  * The S-C-t key combination doesn't get passed to Emacs on X.  The VM mail package, which is not part of Emacs, sometimes does
947      (standard-display-european t)
948    That should be changed to
949      (standard-display-european 1 t)
950    
951  This happens because some X configurations assign the Ctrl-Shift-t  * X runtime problems
 combination the same meaning as the Multi_key.  The offending  
 definition is in the file `...lib/X11/locale/iso8859-1/Compose'; there  
 might be other similar combinations which are grabbed by X for similar  
 purposes.  
952    
953  We think that this can be countermanded with the `xmodmap' utility, if  ** X keyboard problems
 you want to be able to bind one of these key sequences within Emacs.  
954    
955  * On Solaris, CTRL-t is ignored by Emacs when you use  *** You "lose characters" after typing Compose Character key.
 the fr.ISO-8859-15 locale (and maybe other related locales).  
956    
957  You can fix this by editing the file:  This is because the Compose Character key is defined as the keysym
958    Multi_key, and Emacs (seeing that) does the proper X11
959    character-composition processing.  If you don't want your Compose key
960    to do that, you can redefine it with xmodmap.
961    
962          /usr/openwin/lib/locale/iso8859-15/Compose  For example, here's one way to turn it into a Meta key:
963    
964  Near the bottom there is a line that reads:      xmodmap -e "keysym Multi_key = Meta_L"
965    
966          Ctrl<t> <quotedbl> <Y>                  : "\276"        threequarters  If all users at your site of a particular keyboard prefer Meta to
967    Compose, you can make the remapping happen automatically by adding the
968    xmodmap command to the xdm setup script for that display.
969    
970  that should read:  *** Using X Windows, control-shift-leftbutton makes Emacs hang.
971    
972          Ctrl<T> <quotedbl> <Y>                  : "\276"        threequarters  Use the shell command `xset bc' to make the old X Menu package work.
973    
974  Note the lower case <t>.  Changing this line should make C-t work.  *** M-SPC seems to be ignored as input.
975    
976  * Emacs on Digital Unix 4.0 fails to build, giving error message  See if your X server is set up to use this as a command
977       Invalid dimension for the charset-ID 160  for character composition.
978    
979  This is due to a bug or an installation problem in GCC 2.8.0.  *** The S-C-t key combination doesn't get passed to Emacs on X.
 Installing a more recent version of GCC fixes the problem.  
980    
981  * Buffers from `with-output-to-temp-buffer' get set up in Help mode.  This happens because some X configurations assign the Ctrl-Shift-t
982    combination the same meaning as the Multi_key.  The offending
983    definition is in the file `...lib/X11/locale/iso8859-1/Compose'; there
984    might be other similar combinations which are grabbed by X for similar
985    purposes.
986    
987  Changes in Emacs 20.4 to the hooks used by that function cause  We think that this can be countermanded with the `xmodmap' utility, if
988  problems for some packages, specifically BBDB.  See the function's  you want to be able to bind one of these key sequences within Emacs.
 documentation for the hooks involved.  BBDB 2.00.06 fixes the problem.  
989    
990  * Under X, C-v and/or other keys don't work.  *** Under X, C-v and/or other keys don't work.
991    
992  These may have been intercepted by your window manager.  In  These may have been intercepted by your window manager.  In
993  particular, AfterStep 1.6 is reported to steal C-v in its default  particular, AfterStep 1.6 is reported to steal C-v in its default
# Line 1230  configuration.  Various Meta keys are al Line 995  configuration.  Various Meta keys are al
995  configuration of the `feel'.  See the WM's documentation for how to  configuration of the `feel'.  See the WM's documentation for how to
996  change this.  change this.
997    
998  * When using Exceed, fonts sometimes appear too tall.  *** Clicking C-mouse-2 in the scroll bar doesn't split the window.
   
 When the display is set to an Exceed X-server and fonts are specified  
 (either explicitly with the -fn option or implicitly with X resources)  
 then the fonts may appear "too tall".  The actual character sizes are  
 correct but there is too much vertical spacing between rows,  which  
 gives the appearance of "double spacing".  
   
 To prevent this, turn off the Exceed's "automatic font substitution"  
 feature (in the font part of the configuration window).  
   
 * Failure in unexec while dumping emacs on Digital Unix 4.0  
   
 This problem manifests itself as an error message  
   
     unexec: Bad address, writing data section to ...  
   
 The user suspects that this happened because his X libraries  
 were built for an older system version,  
   
     ./configure --x-includes=/usr/include --x-libraries=/usr/shlib  
   
 made the problem go away.  
   
 * No visible display on mips-sgi-irix6.2 when compiling with GCC 2.8.1.  
   
 This problem went away after installing the latest IRIX patches  
 as of 8 Dec 1998.  
   
 The same problem has been reported on Irix 6.3.  
   
 * As of version 20.4, Emacs doesn't work properly if configured for  
 the Motif toolkit and linked against the free LessTif library.  The  
 next Emacs release is expected to work with LessTif.  
   
 * Emacs gives the error, Couldn't find per display information.  
   
 This can result if the X server runs out of memory because Emacs uses  
 a large number of fonts.  On systems where this happens, C-h h is  
 likely to cause it.  
   
 We do not know of a way to prevent the problem.  
   
 * Emacs makes HPUX 11.0 crash.  
999    
1000  This is a bug in HPUX; HPUX patch PHKL_16260 is said to fix it.  This currently doesn't work with scroll-bar widgets (and we don't know
1001    a good way of implementing it with widgets).  If Emacs is configured
1002  * Emacs crashes during dumping on the HPPA machine (HPUX 10.20).  --without-toolkit-scroll-bars, C-mouse-2 on the scroll bar does work.
   
 This seems to be due to a GCC bug; it is fixed in GCC 2.8.1.  
   
 * The Hyperbole package causes *Help* buffers not to be displayed in  
 Help mode due to setting `temp-buffer-show-hook' rather than using  
 `add-hook'.  Using `(add-hook 'temp-buffer-show-hook  
 'help-mode-maybe)' after loading Hyperbole should fix this.  
   
 * Versions of the PSGML package earlier than 1.0.3 (stable) or 1.1.2  
 (alpha) fail to parse DTD files correctly in Emacs 20.3 and later.  
 Here is a patch for psgml-parse.el from PSGML 1.0.1 and, probably,  
 earlier versions.  
   
 --- psgml-parse.el      1998/08/21 19:18:18     1.1  
 +++ psgml-parse.el      1998/08/21 19:20:00  
 @@ -2383,7 +2383,7 @@ (defun sgml-push-to-entity (entity &opti  
        (setq sgml-buffer-parse-state nil))  
      (cond  
       ((stringp entity)                 ; a file name  
 -      (save-excursion (insert-file-contents entity))  
 +      (insert-file-contents entity)  
        (setq default-directory (file-name-directory entity)))  
       ((consp (sgml-entity-text entity)) ; external id?  
        (let* ((extid (sgml-entity-text entity))  
   
 * Emacs 21 freezes when visiting a TeX file with AUC TeX installed.  
1003    
1004  Emacs 21 needs version 10 or later of AUC TeX; upgrading should solve  *** Inability to send an Alt-modified key, when Emacs is communicating
1005  these problems.  directly with an X server.
1006    
1007  * No colors in AUC TeX with Emacs 21.  If you have tried to bind an Alt-modified key as a command, and it
1008    does not work to type the command, the first thing you should check is
1009    whether the key is getting through to Emacs.  To do this, type C-h c
1010    followed by the Alt-modified key.  C-h c should say what kind of event
1011    it read.  If it says it read an Alt-modified key, then make sure you
1012    have made the key binding correctly.
1013    
1014  Upgrade to AUC TeX version 10 or later, and make sure it is  If C-h c reports an event that doesn't have the Alt modifier, it may
1015  byte-compiled with Emacs 21.  be because your X server has no key for the Alt modifier.  The X
1016    server that comes from MIT does not set up the Alt modifier by
1017    default.
1018    
1019  * Running TeX from AUC TeX package with Emacs 20.3 gives a Lisp error  If your keyboard has keys named Alt, you can enable them as follows:
 about a read-only tex output buffer.  
1020    
1021  This problem appeared for AUC TeX version 9.9j and some earlier      xmodmap -e 'add mod2 = Alt_L'
1022  versions.  Here is a patch for the file tex-buf.el in the AUC TeX      xmodmap -e 'add mod2 = Alt_R'
 package.  
   
 diff -c auctex/tex-buf.el~ auctex/tex-buf.el  
 *** auctex/tex-buf.el~  Wed Jul 29 18:35:32 1998  
 --- auctex/tex-buf.el   Sat Sep  5 15:20:38 1998  
 ***************  
 *** 545,551 ****  
         (dir (TeX-master-directory)))  
       (TeX-process-check file)          ; Check that no process is running  
       (setq TeX-command-buffer (current-buffer))  
 !     (with-output-to-temp-buffer buffer)  
       (set-buffer buffer)  
       (if dir (cd dir))  
       (insert "Running `" name "' on `" file "' with ``" command "''\n")  
 - --- 545,552 ----  
         (dir (TeX-master-directory)))  
       (TeX-process-check file)          ; Check that no process is running  
       (setq TeX-command-buffer (current-buffer))  
 !     (let (temp-buffer-show-function temp-buffer-show-hook)  
 !       (with-output-to-temp-buffer buffer))  
       (set-buffer buffer)  
       (if dir (cd dir))  
       (insert "Running `" name "' on `" file "' with ``" command "''\n")  
1023    
1024  * On Irix 6.3, substituting environment variables in file names  If the keyboard has just one key named Alt, then only one of those
1025  in the minibuffer gives peculiar error messages such as  commands is needed.  The modifier `mod2' is a reasonable choice if you
1026    are using an unmodified MIT version of X.  Otherwise, choose any
1027    modifier bit not otherwise used.
1028    
1029     Substituting nonexistent environment variable ""  If your keyboard does not have keys named Alt, you can use some other
1030    keys.  Use the keysym command in xmodmap to turn a function key (or
1031    some other 'spare' key) into Alt_L or into Alt_R, and then use the
1032    commands show above to make them modifier keys.
1033    
1034  This is not an Emacs bug; it is caused by something in SGI patch  Note that if you have Alt keys but no Meta keys, Emacs translates Alt
1035  003082 August 11, 1998.  into Meta.  This is because of the great importance of Meta in Emacs.
1036    
1037  * After a while, Emacs slips into unibyte mode.  ** Window-manager and toolkit-related problems
1038    
1039  The VM mail package, which is not part of Emacs, sometimes does  *** Gnome: Emacs' xterm-mouse-mode doesn't work on the Gnome terminal.
   (standard-display-european t)  
 That should be changed to  
   (standard-display-european 1 t)  
1040    
1041  * Installing Emacs gets an error running `install-info'.  A symptom of this bug is that double-clicks insert a control sequence
1042    into the buffer.  The reason this happens is an apparent
1043    incompatibility of the Gnome terminal with Xterm, which also affects
1044    other programs using the Xterm mouse interface.  A problem report has
1045    been filed.
1046    
1047  You need to install a recent version of Texinfo; that package  *** KDE: When running on KDE, colors or fonts are not as specified for Emacs,
1048  supplies the `install-info' command.  or messed up.
1049    
1050  * Emacs does not recognize the AltGr key, on HPUX.  For example, you could see background you set for Emacs only in the
1051    empty portions of the Emacs display, while characters have some other
1052    background.
1053    
1054  To fix this, set up a file ~/.dt/sessions/sessionetc with executable  This happens because KDE's defaults apply its color and font
1055  rights, containing this text:  definitions even to applications that weren't compiled for KDE.  The
1056    solution is to uncheck the "Apply fonts and colors to non-KDE apps"
1057    option in Preferences->Look&Feel->Style (KDE 2).  In KDE 3, this option
1058    is in the "Colors" section, rather than "Style".
1059    
1060  --------------------------------  Alternatively, if you do want the KDE defaults to apply to other
1061  xmodmap 2> /dev/null - << EOF  applications, but not to Emacs, you could modify the file `Emacs.ad'
1062  keysym Alt_L = Meta_L  (should be in the `/usr/share/apps/kdisplay/app-defaults/' directory)
1063  keysym Alt_R = Meta_R  so that it doesn't set the default background and foreground only for
1064  EOF  Emacs.  For example, make sure the following resources are either not
1065    present or commented out:
1066    
1067  xmodmap - << EOF     Emacs.default.attributeForeground
1068  clear mod1     Emacs.default.attributeBackground
1069  keysym Mode_switch = NoSymbol     Emacs*Foreground
1070  add mod1 = Meta_L     Emacs*Background
 keysym Meta_R = Mode_switch  
 add mod2 = Mode_switch  
 EOF  
 --------------------------------  
1071    
1072  * Emacs hangs on KDE when a large portion of text is killed.  *** KDE: Emacs hangs on KDE when a large portion of text is killed.
1073    
1074  This is caused by a bug in the KDE applet `klipper' which periodically  This is caused by a bug in the KDE applet `klipper' which periodically
1075  requests the X clipboard contents from applications.  Early versions  requests the X clipboard contents from applications.  Early versions
# Line 1394  while, Emacs will print a message: Line 1081  while, Emacs will print a message:
1081    
1082  A workaround is to not use `klipper'.  A workaround is to not use `klipper'.
1083    
1084  * Emacs compiled with DJGPP for MS-DOS/MS-Windows cannot access files  *** CDE: Frames may cover dialogs they created when using CDE.
 in the directory with the special name `dev' under the root of any  
 drive, e.g. `c:/dev'.  
   
 This is an unfortunate side-effect of the support for Unix-style  
 device names such as /dev/null in the DJGPP runtime library.  A  
 work-around is to rename the problem directory to another name.  
   
 * M-SPC seems to be ignored as input.  
   
 See if your X server is set up to use this as a command  
 for character composition.  
   
 * Emacs startup on GNU/Linux systems (and possibly other systems) is slow.  
1085    
1086  This can happen if the system is misconfigured and Emacs can't get the  This can happen if you have "Allow Primary Windows On Top" enabled which
1087  full qualified domain name, FQDN.  You should have your FQDN in the  seems to be the default in the Common Desktop Environment.
1088  /etc/hosts file, something like this:  To change, go in to "Desktop Controls" -> "Window Style Manager"
1089    and uncheck "Allow Primary Windows On Top".
 127.0.0.1       localhost  
 129.187.137.82  nuc04.t30.physik.tu-muenchen.de nuc04  
1090    
1091  The way to set this up may vary on non-GNU systems.  *** Xaw3d : When using Xaw3d scroll bars without arrows, the very first mouse
1092    click in a scroll bar might be ignored by the scroll bar widget.  This
1093    is probably a bug in Xaw3d; when Xaw3d is compiled with arrows, the
1094    problem disappears.
1095    
1096  * Garbled display on non-X terminals when Emacs runs on Digital Unix 4.0.  *** Xaw: There are known binary incompatibilities between Xaw, Xaw3d, neXtaw,
1097    XawM and the few other derivatives of Xaw.  So when you compile with
1098    one of these, it may not work to dynamically link with another one.
1099    For example, strange problems, such as Emacs exiting when you type
1100    "C-x 1", were reported when Emacs compiled with Xaw3d and libXaw was
1101    used with neXtaw at run time.
1102    
1103  So far it appears that running `tset' triggers this problem (when TERM  The solution is to rebuild Emacs with the toolkit version you actually
1104  is vt100, at least).  If you do not run `tset', then Emacs displays  want to use, or set LD_PRELOAD to preload the same toolkit version you
1105  properly.  If someone can tell us precisely which effect of running  built Emacs with.
 `tset' actually causes the problem, we may be able to implement a fix  
 in Emacs.  
1106    
1107  * When you run Ispell from Emacs, it reports a "misalignment" error.  *** Open Motif: Problems with file dialogs in Emacs built with Open Motif.
1108    
1109  This can happen if you compiled the Ispell program to use ASCII  When Emacs 21 is built with Open Motif 2.1, it can happen that the
1110  characters only and then try to use it from Emacs with non-ASCII  graphical file dialog boxes do not work properly.  The "OK", "Filter"
1111  characters, like Latin-1.  The solution is to recompile Ispell with  and "Cancel" buttons do not respond to mouse clicks.  Dragging the
1112  support for 8-bit characters.  file dialog window usually causes the buttons to work again.
1113    
1114  To see whether your Ispell program supports 8-bit characters, type  The solution is to use LessTif instead.  LessTif is a free replacement
1115  this at your shell's prompt:  for Motif.  See the file INSTALL for information on how to do this.
1116    
1117       ispell -vv  Another workaround is not to use the mouse to trigger file prompts,
1118    but to use the keyboard.  This way, you will be prompted for a file in
1119    the minibuffer instead of a graphical file dialog.
1120    
1121  and look in the output for the string "NO8BIT".  If Ispell says  *** LessTif: Problems in Emacs built with LessTif.
 "!NO8BIT (8BIT)", your speller supports 8-bit characters; otherwise it  
 does not.  
1122    
1123  To rebuild Ispell with 8-bit character support, edit the local.h file  The problems seem to depend on the version of LessTif and the Motif
1124  in the Ispell distribution and make sure it does _not_ define NO8BIT.  emulation for which it is set up.
 Then rebuild the speller.  
1125    
1126  Another possible cause for "misalignment" error messages is that the  Only the Motif 1.2 emulation seems to be stable enough in LessTif.
1127  version of Ispell installed on your machine is old.  Upgrade.  Lesstif 0.92-17's Motif 1.2 emulation seems to work okay on FreeBSD.
1128    On GNU/Linux systems, lesstif-0.92.6 configured with "./configure
1129    --enable-build-12 --enable-default-12" is reported to be the most
1130    successful.  The binary GNU/Linux package
1131    lesstif-devel-0.92.0-1.i386.rpm was reported to have problems with
1132    menu placement.
1133    
1134  Yet another possibility is that you are trying to spell-check a word  On some systems, even with Motif 1.2 emulation, Emacs occasionally
1135  in a language that doesn't fit the dictionary you choose for use by  locks up, grabbing all mouse and keyboard events.  We still don't know
1136  Ispell.  (Ispell can only spell-check one language at a time, because  what causes these problems; they are not reproducible by Emacs
1137  it uses a single dictionary.)  Make sure that the text you are  developers.
 spelling and the dictionary used by Ispell conform to each other.  
1138    
1139  If your spell-checking program is Aspell, it has been reported that if  *** Motif: The Motif version of Emacs paints the screen a solid color.
 you have a personal configuration file (normally ~/.aspell.conf), it  
 can cause this error.  Remove that file, execute `ispell-kill-ispell'  
 in Emacs, and then try spell-checking again.  
1140    
1141  * On Linux-based GNU systems using libc versions 5.4.19 through  This has been observed to result from the following X resource:
 5.4.22, Emacs crashes at startup with a segmentation fault.  
1142    
1143  This problem happens if libc defines the symbol __malloc_initialized.     Emacs*default.attributeFont: -*-courier-medium-r-*-*-*-140-*-*-*-*-iso8859-*
 One known solution is to upgrade to a newer libc version.  5.4.33 is  
 known to work.  
1144    
1145  * On MS-Windows, you cannot use the right-hand ALT key and the left-hand  That the resource has this effect indicates a bug in something, but we
1146  CTRL key together to type a Control-Meta character.  do not yet know what.  If it is an Emacs bug, we hope someone can
1147    explain what the bug is so we can fix it.  In the mean time, removing
1148    the resource prevents the problem.
1149    
1150  This is a consequence of a misfeature beyond Emacs's control.  ** General X problems
1151    
1152  Under Windows, the AltGr key on international keyboards generates key  *** Redisplay using X11 is much slower than previous Emacs versions.
 events with the modifiers Right-Alt and Left-Ctrl.  Since Emacs cannot  
 distinguish AltGr from an explicit Right-Alt and Left-Ctrl  
 combination, whenever it sees Right-Alt and Left-Ctrl it assumes that  
 AltGr has been pressed.  The variable `w32-recognize-altgr' can be set  
 to nil to tell Emacs that AltGr is really Ctrl and Alt.  
1153    
1154  * Emacs crashes when using the Exceed 6.0 X server  We've noticed that certain X servers draw the text much slower when
1155    scroll bars are on the left.  We don't know why this happens.  If this
1156    happens to you, you can work around it by putting the scroll bars
1157    on the right (as they were in Emacs 19).
1158    
1159  If you are using Exceed 6.1, upgrade to a later version.  This was  Here's how to do this:
 reported to prevent the crashes.  
1160    
1161  * Under some X-servers running on MS-Windows, Emacs' display is incorrect    (set-scroll-bar-mode 'right)
1162    
1163  The symptoms are that Emacs does not completely erase blank areas of the  If you're not sure whether (or how much) this problem affects you,
1164  screen during scrolling or some other screen operations (e.g., selective  try that and see how much difference it makes.  To set things back
1165  display or when killing a region).  M-x recenter will cause the screen  to normal, do
 to be completely redisplayed and the "extra" characters will disappear.  
1166    
1167  This is known to occur under Exceed 6, and possibly earlier versions    (set-scroll-bar-mode 'left)
 as well; it is reportedly solved in version 6.2.0.16 and later.  The  
 problem lies in the X-server settings.  
1168    
1169  There are reports that you can solve the problem with Exceed by  *** Error messages about undefined colors on X.
 running `Xconfig' from within NT, choosing "X selection", then  
 un-checking the boxes "auto-copy X selection" and "auto-paste to X  
 selection".  
1170    
1171  Of this does not work, please inform bug-gnu-emacs@gnu.org.  Then  The messages might say something like this:
 please call support for your X-server and see if you can get a fix.  
 If you do, please send it to bug-gnu-emacs@gnu.org so we can list it  
 here.  
1172    
1173  * On Solaris 2, Emacs dumps core when built with Motif.     Unable to load color "grey95"
1174    
1175  The Solaris Motif libraries are buggy, at least up through Solaris 2.5.1.  (typically, in the `*Messages*' buffer), or something like this:
 Install the current Motif runtime library patch appropriate for your host.  
 (Make sure the patch is current; some older patch versions still have the bug.)  
 You should install the other patches recommended by Sun for your host, too.  
 You can obtain Sun patches from ftp://sunsolve.sun.com/pub/patches/;  
 look for files with names ending in `.PatchReport' to see which patches  
 are currently recommended for your host.  
1176    
1177  On Solaris 2.6, Emacs is said to work with Motif when Solaris patch    Error while displaying tooltip: (error Undefined color lightyellow)
 105284-12 is installed, but fail when 105284-15 is installed.  
 105284-18 might fix it again.  
1178    
1179  * On Solaris 2.6 and 7, the Compose key does not work.  These problems could happen if some other X program has used up too
1180    many colors of the X palette, leaving Emacs with insufficient system
1181    resources to load all the colors it needs.
1182    
1183  This is a bug in Motif in Solaris.  Supposedly it has been fixed for  A solution is to exit the offending X programs before starting Emacs.
 the next major release of Solaris.  However, if someone with Sun  
 support complains to Sun about the bug, they may release a patch.  
 If you do this, mention Sun bug #4188711.  
1184    
1185  One workaround is to use a locale that allows non-ASCII characters.  *** Improving performance with slow X connections.
 For example, before invoking emacs, set the LC_ALL environment  
 variable to "en_US" (American English).  The directory /usr/lib/locale  
 lists the supported locales; any locale other than "C" or "POSIX"  
 should do.  
1186    
1187  pen@lysator.liu.se says (Feb 1998) that the Compose key does work  There are several ways to improve this performance, any subset of which can
1188  if you link with the MIT X11 libraries instead of the Solaris X11  be carried out at the same time:
 libraries.  
1189    
1190  * Frames may cover dialogs they created when using CDE.  1) If you don't need X Input Methods (XIM) for entering text in some
1191       language you use, you can improve performance on WAN links by using
1192       the X resource useXIM to turn off use of XIM.  This does not affect
1193       the use of Emacs' own input methods, which are part of the Leim
1194       package.
1195    
1196  This can happen if you have "Allow Primary Windows On Top" enabled which  2) If the connection is very slow, you might also want to consider
1197  seems to be the default in the Common Desktop Environment.     switching off scroll bars, menu bar, and tool bar.
 To change, go in to "Desktop Controls" -> "Window Style Manager"  
 and uncheck "Allow Primary Windows On Top".  
1198    
1199  * Emacs does not know your host's fully-qualified domain name.  3) Use ssh to forward the X connection, and enable compression on this
1200       forwarded X connection (ssh -XC remotehostname emacs ...).
1201    
1202  You need to configure your machine with a fully qualified domain name,  4) Use lbxproxy on the remote end of the connection.  This is an interface
1203  either in /etc/hosts, /etc/hostname, the NIS, or wherever your system     to the low bandwidth X extension in most modern X servers, which
1204  calls for specifying this.     improves performance dramatically, at the slight expense of correctness
1205       of the X protocol.  lbxproxy acheives the performance gain by grouping
1206       several X requests in one TCP packet and sending them off together,
1207       instead of requiring a round-trip for each X request in a seperate
1208       packet.  The switches that seem to work best for emacs are:
1209        -noatomsfile  -nowinattr  -cheaterrors -cheatevents
1210       Note that the -nograbcmap option is known to cause problems.
1211       For more about lbxproxy, see:
1212       http://www.xfree86.org/4.3.0/lbxproxy.1.html
1213    
1214  If you cannot fix the configuration, you can set the Lisp variable  *** Emacs gives the error, Couldn't find per display information.
 mail-host-address to the value you want.  
1215    
1216  * Error 12 (virtual memory exceeded) when dumping Emacs, on UnixWare 2.1  This can result if the X server runs out of memory because Emacs uses
1217    a large number of fonts.  On systems where this happens, C-h h is
1218    likely to cause it.
1219    
1220  Paul Abrahams (abrahams@acm.org) reports that with the installed  We do not know of a way to prevent the problem.
 virtual memory settings for UnixWare 2.1.2, an Error 12 occurs during  
 the "make" that builds Emacs, when running temacs to dump emacs.  That  
 error indicates that the per-process virtual memory limit has been  
 exceeded.  The default limit is probably 32MB.  Raising the virtual  
 memory limit to 40MB should make it possible to finish building Emacs.  
1221    
1222  You can do this with the command `ulimit' (sh) or `limit' (csh).  *** Emacs does not notice when you release the mouse.
 But you have to be root to do it.  
1223    
1224  According to Martin Sohnius, you can also retune this in the kernel:  There are reports that this happened with (some) Microsoft mice and
1225    that replacing the mouse made it stop.
1226    
1227      # /etc/conf/bin/idtune SDATLIM 33554432         ## soft data size limit  *** You can't select from submenus (in the X toolkit version).
     # /etc/conf/bin/idtune HDATLIM 33554432         ## hard "  
     # /etc/conf/bin/idtune SVMMSIZE unlimited       ## soft process size limit  
     # /etc/conf/bin/idtune HVMMSIZE unlimited       ## hard "  
     # /etc/conf/bin/idbuild -B  
1228    
1229  (He recommends you not change the stack limit, though.)  On certain systems, mouse-tracking and selection in top-level menus
1230  These changes take effect when you reboot.  works properly with the X toolkit, but neither of them works when you
1231    bring up a submenu (such as Bookmarks or Compare or Apply Patch, in
1232    the Files menu).
1233    
1234  * Redisplay using X11 is much slower than previous Emacs versions.  This works on most systems.  There is speculation that the failure is
1235    due to bugs in old versions of X toolkit libraries, but no one really
1236    knows.  If someone debugs this and finds the precise cause, perhaps a
1237    workaround can be found.
1238    
1239  We've noticed that certain X servers draw the text much slower when  *** An error message such as `X protocol error: BadMatch (invalid
1240  scroll bars are on the left.  We don't know why this happens.  If this  parameter attributes) on protocol request 93'.
 happens to you, you can work around it by putting the scroll bars  
 on the right (as they were in Emacs 19).  
1241    
1242  Here's how to do this:  This comes from having an invalid X resource, such as
1243       emacs*Cursor:   black
1244    (which is invalid because it specifies a color name for something
1245    that isn't a color.)
1246    
1247    (set-scroll-bar-mode 'right)  The fix is to correct your X resources.
1248    
1249  If you're not sure whether (or how much) this problem affects you,  *** Slow startup on X11R6 with X windows.
 try that and see how much difference it makes.  To set things back  
 to normal, do  
1250    
1251    (set-scroll-bar-mode 'left)  If Emacs takes two minutes to start up on X11R6, see if your X
1252    resources specify any Adobe fonts.  That causes the type-1 font
1253    renderer to start up, even if the font you asked for is not a type-1
1254    font.
1255    
1256  * Under X11, some characters appear as hollow boxes.  One way to avoid this problem is to eliminate the type-1 fonts from
1257    your font path, like this:
1258    
1259  Each X11 font covers just a fraction of the characters that Emacs          xset -fp /usr/X11R6/lib/X11/fonts/Type1/
 supports.  To display the whole range of Emacs characters requires  
 many different fonts, collected into a fontset.  
1260    
1261  If some of the fonts called for in your fontset do not exist on your X  *** Pull-down menus appear in the wrong place, in the toolkit version of Emacs.
 server, then the characters that have no font appear as hollow boxes.  
 You can remedy the problem by installing additional fonts.  
1262    
1263  The intlfonts distribution includes a full spectrum of fonts that can  An X resource of this form can cause the problem:
 display all the characters Emacs supports.  
1264    
1265  Another cause of this for specific characters is fonts which have a     Emacs*geometry:      80x55+0+0
 missing glyph and no default character.  This is known ot occur for  
 character number 160 (no-break space) in some fonts, such as Lucida  
 but Emacs sets the display table for the unibyte and Latin-1 version  
 of this character to display a space.  
1266    
1267  * Under X11, some characters appear improperly aligned in their lines.  This resource is supposed to apply, and does apply, to the menus
1268    individually as well as to Emacs frames.  If that is not what you
1269    want, rewrite the resource.
1270    
1271  You may have bad X11 fonts; try installing the intlfonts distribution.  To check thoroughly for such resource specifications, use `xrdb
1272    -query' to see what resources the X server records, and also look at
1273    the user's ~/.Xdefaults and ~/.Xdefaults-* files.
1274    
1275  * Certain fonts make each line take one pixel more than it "should".  *** --with-x-toolkit version crashes when used with shared libraries.
1276    
1277  This is because these fonts contain characters a little taller  On some systems, including Sunos 4 and DGUX 5.4.2 and perhaps others,
1278  than the font's nominal height.  Emacs needs to make sure that  unexec doesn't work properly with the shared library for the X
1279  lines do not overlap.  toolkit.  You might be able to work around this by using a nonshared
1280    libXt.a library.  The real fix is to upgrade the various versions of
1281    unexec and/or ralloc.  We think this has been fixed on Sunos 4
1282    and Solaris in version 19.29.
1283    
1284  * You request inverse video, and the first Emacs frame is in inverse  *** Emacs running under X Windows does not handle mouse clicks.
1285  video, but later frames are not in inverse video.  *** `emacs -geometry 80x20' finds a file named `80x20'.
1286    
1287  This can happen if you have an old version of the custom library in  One cause of such problems is having (setq term-file-prefix nil) in
1288  your search path for Lisp packages.  Use M-x list-load-path-shadows to  your .emacs file.  Another cause is a bad value of EMACSLOADPATH in
1289  check whether this is true.  If it is, delete the old custom library.  the environment.
1290    
1291  * In FreeBSD 2.1.5, useless symbolic links remain in /tmp or other  *** Emacs fails to get default settings from X Windows server.
 directories that have the +t bit.  
1292    
1293  This is because of a kernel bug in FreeBSD 2.1.5 (fixed in 2.2).  The X library in X11R4 has a bug; it interchanges the 2nd and 3rd
1294  Emacs uses symbolic links to implement file locks.  In a directory  arguments to XGetDefaults.  Define the macro XBACKWARDS in config.h to
1295  with +t bit, the directory owner becomes the owner of the symbolic  tell Emacs to compensate for this.
 link, so that it cannot be removed by anyone else.  
1296    
1297  If you don't like those useless links, you can let Emacs not to using  I don't believe there is any way Emacs can determine for itself
1298  file lock by adding #undef CLASH_DETECTION to config.h.  whether this problem is present on a given system.
1299    
1300  * When using M-x dbx with the SparcWorks debugger, the `up' and `down'  *** X Windows doesn't work if DISPLAY uses a hostname.
 commands do not move the arrow in Emacs.  
1301    
1302  You can fix this by adding the following line to `~/.dbxinit':  People have reported kernel bugs in certain systems that cause Emacs
1303    not to work with X Windows if DISPLAY is set using a host name.  But
1304    the problem does not occur if DISPLAY is set to `unix:0.0'.  I think
1305    the bug has to do with SIGIO or FIONREAD.
1306    
1307   dbxenv output_short_file_name off  You may be able to compensate for the bug by doing (set-input-mode nil nil).
1308    However, that has the disadvantage of turning off interrupts, so that
1309    you are unable to quit out of a Lisp program by typing C-g.
1310    
1311  * Emacs says it has saved a file, but the file does not actually  The easy way to do this is to put
 appear on disk.  
1312    
1313  This can happen on certain systems when you are using NFS, if the    (setq x-sigio-bug t)
 remote disk is full.  It is due to a bug in NFS (or certain NFS  
 implementations), and there is apparently nothing Emacs can do to  
 detect the problem.  Emacs checks the failure codes of all the system  
 calls involved in writing a file, including `close'; but in the case  
 where the problem occurs, none of those system calls fails.  
1314    
1315  * "Compose Character" key does strange things when used as a Meta key.  in your site-init.el file.
1316    
1317  If you define one key to serve as both Meta and Compose Character, you  * Runtime problems on character termunals
 will get strange results.  In previous Emacs versions, this "worked"  
 in that the key acted as Meta--that's because the older Emacs versions  
 did not try to support Compose Character.  Now Emacs tries to do  
 character composition in the standard X way.  This means that you  
 must pick one meaning or the other for any given key.  
1318    
1319  You can use both functions (Meta, and Compose Character) if you assign  ** Emacs spontaneously displays "I-search: " at the bottom of the screen.
 them to two different keys.  
1320    
1321  * Emacs gets a segmentation fault at startup, on AIX4.2.  This means that Control-S/Control-Q (XON/XOFF) "flow control" is being
1322    used.  C-s/C-q flow control is bad for Emacs editors because it takes
1323    away C-s and C-q as user commands.  Since editors do not output long
1324    streams of text without user commands, there is no need for a
1325    user-issuable "stop output" command in an editor; therefore, a
1326    properly designed flow control mechanism would transmit all possible
1327    input characters without interference.  Designing such a mechanism is
1328    easy, for a person with at least half a brain.
1329    
1330  If you are using IBM's xlc compiler, compile emacs.c  There are three possible reasons why flow control could be taking place:
 without optimization; that should avoid the problem.  
1331    
1332  * movemail compiled with POP support can't connect to the POP server.    1) Terminal has not been told to disable flow control
1333      2) Insufficient padding for the terminal in use
1334      3) Some sort of terminal concentrator or line switch is responsible
1335    
1336  Make sure that the `pop' entry in /etc/services, or in the services  First of all, many terminals have a set-up mode which controls whether
1337  NIS map if your machine uses NIS, has the same port number as the  they generate XON/XOFF flow control characters.  This must be set to
1338  entry on the POP server.  A common error is for the POP server to be  "no XON/XOFF" in order for Emacs to work.  Sometimes there is an
1339  listening on port 110, the assigned port for the POP3 protocol, while  escape sequence that the computer can send to turn flow control off
1340  the client is trying to connect on port 109, the assigned port for the  and on.  If so, perhaps the termcap `ti' string should turn flow
1341  old POP protocol.  control off, and the `te' string should turn it on.
1342    
1343  * Emacs crashes in x-popup-dialog.  Once the terminal has been told "no flow control", you may find it
1344    needs more padding.  The amount of padding Emacs sends is controlled
1345    by the termcap entry for the terminal in use, and by the output baud
1346    rate as known by the kernel.  The shell command `stty' will print
1347    your output baud rate; `stty' with suitable arguments will set it if
1348    it is wrong.  Setting to a higher speed causes increased padding.  If
1349    the results are wrong for the correct speed, there is probably a
1350    problem in the termcap entry.  You must speak to a local Unix wizard
1351    to fix this.  Perhaps you are just using the wrong terminal type.
1352    
1353  This can happen if the dialog widget cannot find the font it wants to  For terminals that lack a "no flow control" mode, sometimes just
1354  use.  You can work around the problem by specifying another font with  giving lots of padding will prevent actual generation of flow control
1355  an X resource--for example, `Emacs.dialog*.font: 9x15' (or any font that  codes.  You might as well try it.
 happens to exist on your X server).  
1356    
1357  * Emacs crashes when you use Bibtex mode.  If you are really unlucky, your terminal is connected to the computer
1358    through a concentrator which sends XON/XOFF flow control to the
1359    computer, or it insists on sending flow control itself no matter how
1360    much padding you give it.  Unless you can figure out how to turn flow
1361    control off on this concentrator (again, refer to your local wizard),
1362    you are screwed!  You should have the terminal or concentrator
1363    replaced with a properly designed one.  In the mean time, some drastic
1364    measures can make Emacs semi-work.
1365    
1366  This happens if your system puts a small limit on stack size.  You can  You can make Emacs ignore C-s and C-q and let the operating system
1367  prevent the problem by using a suitable shell command (often `ulimit')  handle them.  To do this on a per-session basis, just type M-x
1368  to raise the stack size limit before you run Emacs.  enable-flow-control RET.  You will see a message that C-\ and C-^ are
1369    now translated to C-s and C-q.  (Use the same command M-x
1370    enable-flow-control to turn *off* this special mode.  It toggles flow
1371    control handling.)
1372    
1373  Patches to raise the stack size limit automatically in `main'  If C-\ and C-^ are inconvenient for you (for example, if one of them
1374  (src/emacs.c) on various systems would be greatly appreciated.  is the escape character of your terminal concentrator), you can choose
1375    other characters by setting the variables flow-control-c-s-replacement
1376    and flow-control-c-q-replacement.  But choose carefully, since all
1377    other control characters are already used by emacs.
1378    
1379  * Emacs crashes with SIGBUS or SIGSEGV on HPUX 9 after you delete a frame.  IMPORTANT: if you type C-s by accident while flow control is enabled,
1380    Emacs output will freeze, and you will have to remember to type C-q in
1381    order to continue.
1382    
1383  We think this is due to a bug in the X libraries provided by HP.  With  If you work in an environment where a majority of terminals of a
1384  the alternative X libraries in /usr/contrib/mitX11R5/lib, the problem  certain type are flow control hobbled, you can use the function
1385  does not happen.  `enable-flow-control-on' to turn on this flow control avoidance scheme
1386    automatically.  Here is an example:
1387    
1388  * Emacs crashes with SIGBUS or SIGSEGV on Solaris after you delete a frame.  (enable-flow-control-on "vt200" "vt300" "vt101" "vt131")
1389    
1390  We suspect that this is a similar bug in the X libraries provided by  If this isn't quite correct (e.g. you have a mixture of flow-control hobbled
1391  Sun.  There is a report that one of these patches fixes the bug and  and good vt200 terminals), you can still run enable-flow-control
1392  makes the problem stop:  manually.
1393    
1394  105216-01 105393-01 105518-01 105621-01 105665-01 105615-02 105216-02  I have no intention of ever redesigning the Emacs command set for the
1395  105667-01 105401-08 105615-03 105621-02 105686-02 105736-01 105755-03  assumption that terminals use C-s/C-q flow control.  XON/XOFF flow
1396  106033-01 105379-01 105786-01 105181-04 105379-03 105786-04 105845-01  control technique is a bad design, and terminals that need it are bad
1397  105284-05 105669-02 105837-01 105837-02 105558-01 106125-02 105407-01  merchandise and should not be purchased.  Now that X is becoming
1398    widespread, XON/XOFF seems to be on the way out.  If you can get some
1399    use out of GNU Emacs on inferior terminals, more power to you, but I
1400    will not make Emacs worse for properly designed systems for the sake
1401    of inferior systems.
1402    
1403  Another person using a newer system (kernel patch level Generic_105181-06)  ** Control-S and Control-Q commands are ignored completely.
 suspects that the bug was fixed by one of these more recent patches:  
1404    
1405  106040-07  SunOS 5.6: X Input & Output Method patch  For some reason, your system is using brain-damaged C-s/C-q flow
1406  106222-01  OpenWindows 3.6: filemgr (ff.core) fixes  control despite Emacs's attempts to turn it off.  Perhaps your
1407  105284-12  Motif 1.2.7: sparc Runtime library patch  terminal is connected to the computer through a concentrator
1408    that wants to use flow control.
1409    
1410  * Problems running Perl under Emacs on MS-Windows NT/95.  You should first try to tell the concentrator not to use flow control.
1411    If you succeed in this, try making the terminal work without
1412    flow control, as described in the preceding section.
1413    
1414  `perl -de 0' just hangs when executed in an Emacs subshell.  If that line of approach is not successful, map some other characters
1415  The fault lies with Perl (indirectly with Windows NT/95).  into C-s and C-q using keyboard-translate-table.  The example above
1416    shows how to do this with C-^ and C-\.
1417    
1418  The problem is that the Perl debugger explicitly opens a connection to  ** Screen is updated wrong, but only on one kind of terminal.
 "CON", which is the DOS/NT equivalent of "/dev/tty", for interacting  
 with the user.  
1419    
1420  On Unix, this is okay, because Emacs (or the shell?) creates a  This could mean that the termcap entry you are using for that
1421  pseudo-tty so that /dev/tty is really the pipe Emacs is using to  terminal is wrong, or it could mean that Emacs has a bug handing
1422  communicate with the subprocess.  the combination of features specified for that terminal.
1423    
1424  On NT, this fails because CON always refers to the handle for the  The first step in tracking this down is to record what characters
1425  relevant console (approximately equivalent to a tty), and cannot be  Emacs is sending to the terminal.  Execute the Lisp expression
1426  redirected to refer to the pipe Emacs assigned to the subprocess as  (open-termscript "./emacs-script") to make Emacs write all
1427  stdin.  terminal output into the file ~/emacs-script as well; then do
1428    what makes the screen update wrong, and look at the file
1429    and decode the characters using the manual for the terminal.
1430    There are several possibilities:
1431    
1432  A workaround is to modify perldb.pl to use STDIN/STDOUT instead of CON.  1) The characters sent are correct, according to the terminal manual.
1433    
1434  For Perl 4:  In this case, there is no obvious bug in Emacs, and most likely you
1435    need more padding, or possibly the terminal manual is wrong.
1436    
1437      *** PERL/LIB/PERLDB.PL.orig Wed May 26 08:24:18 1993  2) The characters sent are incorrect, due to an obscure aspect
1438      --- PERL/LIB/PERLDB.PL      Mon Jul 01 15:28:16 1996   of the terminal behavior not described in an obvious way
1439      ***************   by termcap.
     *** 68,74 ****  
           $rcfile=".perldb";  
       }  
       else {  
     !     $console = "con";  
           $rcfile="perldb.ini";  
       }  
1440    
1441      --- 68,74 ----  This case is hard.  It will be necessary to think of a way for
1442            $rcfile=".perldb";  Emacs to distinguish between terminals with this kind of behavior
1443        }  and other terminals that behave subtly differently but are
1444        else {  classified the same by termcap; or else find an algorithm for
1445      !     $console = "";  Emacs to use that avoids the difference.  Such changes must be
1446            $rcfile="perldb.ini";  tested on many kinds of terminals.
       }  
1447    
1448    3) The termcap entry is wrong.
1449    
1450      For Perl 5:  See the file etc/TERMS for information on changes
1451      *** perl/5.001/lib/perl5db.pl.orig  Sun Jun 04 21:13:40 1995  that are known to be needed in commonly used termcap entries
1452      --- perl/5.001/lib/perl5db.pl       Mon Jul 01 17:00:08 1996  for certain terminals.
     ***************  
     *** 22,28 ****  
           $rcfile=".perldb";  
       }  
       elsif (-e "con") {  
     !     $console = "con";  
           $rcfile="perldb.ini";  
       }  
       else {  
     --- 22,28 ----  
           $rcfile=".perldb";  
       }  
       elsif (-e "con") {  
     !     $console = "";  
           $rcfile="perldb.ini";  
       }  
       else {  
1453    
1454  * Problems on MS-DOG if DJGPP v2.0 is used to compile Emacs:  4) The characters sent are incorrect, and clearly cannot be
1455     right for any terminal with the termcap entry you were using.
1456    
1457  There are two DJGPP library bugs which cause problems:  This is unambiguously an Emacs bug, and can probably be fixed
1458    in termcap.c, tparam.c, term.c, scroll.c, cm.c or dispnew.c.
1459    
1460    * Running `shell-command' (or `compile', or `grep') you get  ** Control-S and Control-Q commands are ignored completely on a net connection.
     `Searching for program: permission denied (EACCES), c:/command.com';  
   * After you shell to DOS, Ctrl-Break kills Emacs.  
1461    
1462  To work around these bugs, you can use two files in the msdos  Some versions of rlogin (and possibly telnet) do not pass flow
1463  subdirectory: `is_exec.c' and `sigaction.c'.  Compile them and link  control characters to the remote system to which they connect.
1464  them into the Emacs executable `temacs'; then they will replace the  On such systems, emacs on the remote system cannot disable flow
1465  incorrect library functions.  control on the local system.
1466    
1467  * When compiling with DJGPP on MS-Windows NT, "config msdos" fails.  One way to cure this is to disable flow control on the local host
1468    (the one running rlogin, not the one running rlogind) using the
1469    stty command, before starting the rlogin process.  On many systems,
1470    "stty start u stop u" will do this.
1471    
1472  If the error message is "VDM has been already loaded", this is because  Some versions of tcsh will prevent even this from working.  One way
1473  Windows has a program called `redir.exe' that is incompatible with a  around this is to start another shell before starting rlogin, and
1474  program by the same name supplied with DJGPP, which is used by  issue the stty command to disable flow control from that shell.
 config.bat.  To resolve this, move the DJGPP's `bin' subdirectory to  
 the front of your PATH environment variable.  
1475    
1476  * When compiling with DJGPP on MS-Windows 95, Make fails for some targets  If none of these methods work, the best solution is to type
1477  like make-docfile.  M-x enable-flow-control at the beginning of your emacs session, or
1478    if you expect the problem to continue, add a line such as the
1479    following to your .emacs (on the host running rlogind):
1480    
1481  This can happen if long file name support (the setting of environment  (enable-flow-control-on "vt200" "vt300" "vt101" "vt131")
 variable LFN) when Emacs distribution was unpacked and during  
 compilation are not the same.  See the MSDOG section of INSTALL for  
 the explanation of how to avoid this problem.  
1482    
1483  * Emacs compiled for MSDOS cannot find some Lisp files, or other  See the entry about spontaneous display of I-search (above) for more
1484  run-time support files, when long filename support is enabled.  info.
1485    
1486  Usually, this problem will manifest itself when Emacs exits  ** Output from Control-V is slow.
 immediately after flashing the startup screen, because it cannot find  
 the Lisp files it needs to load at startup.  Redirect Emacs stdout  
 and stderr to a file to see the error message printed by Emacs.  
1487    
1488  Another manifestation of this problem is that Emacs is unable to load  On many bit-map terminals, scrolling operations are fairly slow.
1489  the support for editing program sources in languages such as C and  Often the termcap entry for the type of terminal in use fails
1490  Lisp.  to inform Emacs of this.  The two lines at the bottom of the screen
1491    before a Control-V command are supposed to appear at the top after
1492    the Control-V command.  If Emacs thinks scrolling the lines is fast,
1493    it will scroll them to the top of the screen.
1494    
1495  This can happen if the Emacs distribution was unzipped without LFN  If scrolling is slow but Emacs thinks it is fast, the usual reason is
1496  support, thus causing long filenames to be truncated to the first 6  that the termcap entry for the terminal you are using does not
1497  characters and a numeric tail that Windows 95 normally attaches to it.  specify any padding time for the `al' and `dl' strings.  Emacs
1498  You should unzip the files again with a utility that supports long  concludes that these operations take only as much time as it takes to
1499  filenames (such as djtar from DJGPP or InfoZip's UnZip program  send the commands at whatever line speed you are using.  You must
1500  compiled with DJGPP v2).  The MSDOG section of the file INSTALL  fix the termcap entry to specify, for the `al' and `dl', as much
1501  explains this issue in more detail.  time as the operations really take.
1502    
1503  Another possible reason for such failures is that Emacs compiled for  Currently Emacs thinks in terms of serial lines which send characters
1504  MSDOS is used on Windows NT, where long file names are not supported  at a fixed rate, so that any operation which takes time for the
1505  by this version of Emacs, but the distribution was unpacked by an  terminal to execute must also be padded.  With bit-map terminals
1506  unzip program that preserved the long file names instead of truncating  operated across networks, often the network provides some sort of
1507  them to DOS 8+3 limits.  To be useful on NT, the MSDOS port of Emacs  flow control so that padding is never needed no matter how slow
1508  must be unzipped by a DOS utility, so that long file names are  an operation is.  You must still specify a padding time if you want
1509  properly truncated.  Emacs to realize that the operation takes a long time.  This will
1510    cause padding characters to be sent unnecessarily, but they do
1511    not really cost much.  They will be transmitted while the scrolling
1512    is happening and then discarded quickly by the terminal.
1513    
1514  * Emacs compiled with DJGPP complains at startup:  Most bit-map terminals provide commands for inserting or deleting
1515    multiple lines at once.  Define the `AL' and `DL' strings in the
1516    termcap entry to say how to do these things, and you will have
1517    fast output without wasted padding characters.  These strings should
1518    each contain a single %-spec saying how to send the number of lines
1519    to be scrolled.  These %-specs are like those in the termcap
1520    `cm' string.
1521    
1522    "Wrong type of argument: internal-facep, msdos-menu-active-face"  You should also define the `IC' and `DC' strings if your terminal
1523    has a command to insert or delete multiple characters.  These
1524    take the number of positions to insert or delete as an argument.
1525    
1526  This can happen if you define an environment variable `TERM'.  Emacs  A `cs' string to set the scrolling region will reduce the amount
1527  on MSDOS uses an internal terminal emulator which is disabled if the  of motion you see on the screen when part of the screen is scrolled.
 value of `TERM' is anything but the string "internal".  Emacs then  
 works as if its terminal were a dumb glass teletype that doesn't  
 support faces.  To work around this, arrange for `TERM' to be  
 undefined when Emacs runs.  The best way to do that is to add an  
 [emacs] section to the DJGPP.ENV file which defines an empty value for  
 `TERM'; this way, only Emacs gets the empty value, while the rest of  
 your system works as before.  
1528    
1529  * On MS-Windows 95, Alt-f6 does not get through to Emacs.  ** You type Control-H (Backspace) expecting to delete characters.
1530    
1531  This character seems to be trapped by the kernel in Windows 95.  Put `stty dec' in your .login file and your problems will disappear
1532  You can enter M-f6 by typing ESC f6.  after a day or two.
1533    
1534  * Typing Alt-Shift has strange effects on MS-Windows.  The choice of Backspace for erasure was based on confusion, caused by
1535    the fact that backspacing causes erasure (later, when you type another
1536    character) on most display terminals.  But it is a mistake.  Deletion
1537    of text is not the same thing as backspacing followed by failure to
1538    overprint.  I do not wish to propagate this confusion by conforming
1539    to it.
1540    
1541  This combination of keys is a command to change keyboard layout.  If  For this reason, I believe `stty dec' is the right mode to use,
1542  you proceed to type another non-modifier key before you let go of Alt  and I have designed Emacs to go with that.  If there were a thousand
1543  and Shift, the Alt and Shift act as modifiers in the usual way.  A  other control characters, I would define Control-h to delete as well;
1544  more permanent work around is to change it to another key combination,  but there are not very many other control characters, and I think
1545  or disable it in the keyboard control panel.  that providing the most mnemonic possible Help character is more
1546    important than adapting to people who don't use `stty dec'.
1547    
1548  * `tparam' reported as a multiply-defined symbol when linking with ncurses.  If you are obstinate about confusing buggy overprinting with deletion,
1549    you can redefine Backspace in your .emacs file:
1550      (global-set-key "\b" 'delete-backward-char)
1551    You can probably access  help-command  via f1.
1552    
1553  This problem results from an incompatible change in ncurses, in  ** Colors are not available on a tty or in xterm.
 version 1.9.9e approximately.  This version is unable to provide a  
 definition of tparm without also defining tparam.  This is also  
 incompatible with Terminfo; as a result, the Emacs Terminfo support  
 does not work with this version of ncurses.  
1554    
1555  The fix is to install a newer version of ncurses, such as version 4.2.  Emacs 21 supports colors on character terminals and terminal
1556    emulators, but this support relies on the terminfo or termcap database
1557    entry to specify that the display supports color.  Emacs looks at the
1558    "Co" capability for the terminal to find out how many colors are
1559    supported; it should be non-zero to activate the color support within
1560    Emacs.  (Most color terminals support 8 or 16 colors.)  If your system
1561    uses terminfo, the name of the capability equivalent to "Co" is
1562    "colors".
1563    
1564  * Emacs does not start, complaining that it cannot open termcap database file.  In addition to the "Co" capability, Emacs needs the "op" (for
1565    ``original pair'') capability, which tells how to switch the terminal
1566    back to the default foreground and background colors.  Emacs will not
1567    use colors if this capability is not defined.  If your terminal entry
1568    doesn't provide such a capability, try using the ANSI standard escape
1569    sequence \E[00m (that is, define a new termcap/terminfo entry and make
1570    it use your current terminal's entry plus \E[00m for the "op"
1571    capability).
1572    
1573  If your system uses Terminfo rather than termcap (most modern  Finally, the "NC" capability (terminfo name: "ncv") tells Emacs which
1574  systems do), this could happen if the proper version of  attributes cannot be used with colors.  Setting this capability
1575  ncurses is not visible to the Emacs configure script (i.e. it  incorrectly might have the effect of disabling colors; try setting
1576  cannot be found along the usual path the linker looks for  this capability to `0' (zero) and see if that helps.
 libraries). It can happen because your version of ncurses is  
 obsolete, or is available only in form of binaries.  
1577    
1578  The solution is to install an up-to-date version of ncurses in  Emacs uses the database entry for the terminal whose name is the value
1579  the developer's form (header files, static libraries and  of the environment variable TERM.  With `xterm', a common terminal
1580  symbolic links); in some GNU/Linux distributions (e.g. Debian)  entry that supports color is `xterm-color', so setting TERM's value to
1581  it constitutes a separate package.  `xterm-color' might activate the color support on an xterm-compatible
1582    emulator.
1583    
1584  * Strange results from format %d in a few cases, on a Sun.  Beginning with version 21.4, Emacs supports the --color command-line
1585    option which may be used to force Emacs to use one of a few popular
1586    modes for getting colors on a tty.  For example, --color=ansi8 sets up
1587    for using the ANSI-standard escape sequences that support 8 colors.
1588    
1589  Sun compiler version SC3.0 has been found to miscompile part of  Some modes do not use colors unless you turn on the Font-lock mode.
1590  editfns.c.  The workaround is to compile with some other compiler such  Some people have long ago set their `~/.emacs' files to turn on
1591  as GCC.  Font-lock on X only, so they won't see colors on a tty.  The
1592    recommended way of turning on Font-lock is by typing "M-x
1593    global-font-lock-mode RET" or by customizing the variable
1594    `global-font-lock-mode'.
1595    
1596  * Output from subprocess (such as man or diff) is randomly truncated  * Runtime problems specific to individual Unix variants
 on GNU/Linux systems.  
1597    
1598  This is due to a kernel bug which seems to be fixed in Linux version  ** GNU/Linux
 1.3.75.  
1599    
1600  * Error messages `internal facep []' happen on GNU/Linux systems.  *** GNU/Linux: On Linux-based GNU systems using libc versions 5.4.19 through
1601    5.4.22, Emacs crashes at startup with a segmentation fault.
1602    
1603  There is a report that replacing libc.so.5.0.9 with libc.so.5.2.16  This problem happens if libc defines the symbol __malloc_initialized.
1604  caused this to start happening.  People are not sure why, but the  One known solution is to upgrade to a newer libc version.  5.4.33 is
1605  problem seems unlikely to be in Emacs itself.  Some suspect that it  known to work.
 is actually Xlib which won't work with libc.so.5.2.16.  
1606    
1607  Using the old library version is a workaround.  *** GNU/Linux: After upgrading to a newer version of Emacs,
1608    the Meta key stops working.
1609    
1610  * On Solaris, Emacs crashes if you use (display-time).  This was reported to happen on a GNU/Linux system distributed by
1611    Mandrake.  The reason is that the previous version of Emacs was
1612    modified by Mandrake to make the Alt key act as the Meta key, on a
1613    keyboard where the Windows key is the one which produces the Meta
1614    modifier.  A user who started using a newer version of Emacs, which
1615    was not hacked by Mandrake, expected the Alt key to continue to act as
1616    Meta, and was astonished when that didn't happen.
1617    
1618  This can happen if you configure Emacs without specifying the precise  The solution is to find out what key on your keyboard produces the Meta
1619  version of Solaris that you are using.  modifier, and use that key instead.  Try all of the keys to the left
1620    and to the right of the space bar, together with the `x' key, and see
1621    which combination produces "M-x" in the echo area.  You can also use
1622    the `xmodmap' utility to show all the keys which produce a Meta
1623    modifier:
1624    
1625  * Emacs dumps core on startup, on Solaris.           xmodmap -pk | egrep -i "meta|alt"
1626    
1627  Bill Sebok says that the cause of this is Solaris 2.4 vendor patch  A more convenient way of finding out which keys produce a Meta modifier
1628  102303-05, which extends the Solaris linker to deal with the Solaris  is to use the `xkbprint' utility, if it's available on your system:
 Common Desktop Environment's linking needs.  You can fix the problem  
 by removing this patch and installing patch 102049-02 instead.  
 However, that linker version won't work with CDE.  
1629    
1630  Solaris 2.5 comes with a linker that has this bug.  It is reported that if           xkbprint 0:0 /tmp/k.ps
 you install all the latest patches (as of June 1996), the bug is fixed.  
 We suspect the crucial patch is one of these, but we don't know  
 for certain.  
1631    
1632          103093-03: [README] SunOS 5.5: kernel patch (2140557 bytes)  This produces a PostScript file `/tmp/k.ps' with a picture of your
1633          102832-01: [README] OpenWindows 3.5: Xview Jumbo Patch (4181613 bytes)  keyboard; printing that file on a PostScript printer will show what
1634          103242-04: [README] SunOS 5.5: linker patch (595363 bytes)  keys can serve as Meta.
1635    
1636  (One user reports that the bug was fixed by those patches together  The `xkeycaps' also shows a visual representation of the current
1637  with patches 102980-04, 103279-01, 103300-02, and 103468-01.)  keyboard settings.  It also allows to modify them.
1638    
1639  If you can determine which patch does fix the bug, please tell  *** GNU/Linux: low startup on Linux-based GNU systems.
 bug-gnu-emacs@gnu.org.  
1640    
1641  Meanwhile, the GNU linker links Emacs properly on both Solaris 2.4 and  People using systems based on the Linux kernel sometimes report that
1642  Solaris 2.5.  startup takes 10 to 15 seconds longer than `usual'.
1643    
1644  * Emacs dumps core if lisp-complete-symbol is called, on Solaris.  This is because Emacs looks up the host name when it starts.
1645    Normally, this takes negligible time; the extra delay is due to
1646    improper system configuration.  This problem can occur for both
1647    networked and non-networked machines.
1648    
1649  If you compile Emacs with the -fast or -xO4 option with version 3.0.2  Here is how to fix the configuration.  It requires being root.
 of the Sun C compiler, Emacs dumps core when lisp-complete-symbol is  
 called.  The problem does not happen if you compile with GCC.  
1650    
1651  * "Cannot find callback list" messages from dialog boxes on HPUX, in  **** Networked Case.
 Emacs built with Motif.  
1652    
1653  This problem resulted from a bug in GCC 2.4.5.  Newer GCC versions  First, make sure the files `/etc/hosts' and `/etc/host.conf' both
1654  such as 2.7.0 fix the problem.  exist.  The first line in the `/etc/hosts' file should look like this
1655    (replace HOSTNAME with your host name):
1656    
1657  * On Irix 6.0, make tries (and fails) to build a program named unexelfsgi      127.0.0.1      HOSTNAME
1658    
1659  A compiler bug inserts spaces into the string "unexelfsgi . o"  Also make sure that the `/etc/host.conf' files contains the following
1660  in src/Makefile.  Edit src/Makefile, after configure is run,  lines:
 find that string, and take out the spaces.  
1661    
1662  Compiler fixes in Irix 6.0.1 should eliminate this problem.      order hosts, bind
1663        multi on
1664    
1665  * "out of virtual swap space" on Irix 5.3  Any changes, permanent and temporary, to the host name should be
1666    indicated in the `/etc/hosts' file, since it acts a limited local
1667    database of addresses and names (e.g., some SLIP connections
1668    dynamically allocate ip addresses).
1669    
1670  This message occurs when the system runs out of swap space due to too  **** Non-Networked Case.
 many large programs running.  The solution is either to provide more  
 swap space or to reduce the number of large programs being run.  You  
 can check the current status of the swap space by executing the  
 command `swap -l'.  
1671    
1672  You can increase swap space by changing the file /etc/fstab.  Adding a  The solution described in the networked case applies here as well.
1673  line like this:  However, if you never intend to network your machine, you can use a
1674    simpler solution: create an empty `/etc/host.conf' file.  The command
1675    `touch /etc/host.conf' suffices to create the file.  The `/etc/hosts'
1676    file is not necessary with this approach.
1677    
1678  /usr/swap/swap.more     swap    swap    pri=3 0 0  *** GNU/Linux: Emacs on a tty switches the cursor to large blinking block.
1679    
1680  where /usr/swap/swap.more is a file previously created (for instance  This was reported to happen on some GNU/Linux systems which use
1681  by using /etc/mkfile), will increase the swap space by the size of  ncurses version 5.0, but could be relevant for other versions as well.
1682  that file.  Execute `swap -m' or reboot the machine to activate the  These versions of ncurses come with a `linux' terminfo entry, where
1683  new swap area.  See the manpages for `swap' and `fstab' for further  the "cvvis" capability (termcap "vs") is defined as "\E[?25h\E[?8c"
1684  information.  (show cursor, change size).  This escape sequence switches on a
1685    blinking hardware text-mode cursor whose size is a full character
1686    cell.  This blinking cannot be stopped, since a hardware cursor
1687    always blinks.
1688    
1689  The objectserver daemon can use up lots of memory because it can be  A work-around is to redefine the "cvvis" capability so that it
1690  swamped with NIS information.  It collects information about all users  enables a *software* cursor.  The software cursor works by inverting
1691  on the network that can log on to the host.  the colors of the character at point, so what you see is a block
1692    cursor that doesn't blink.  For this to work, you need to redefine
1693    the "cnorm" capability as well, so that it operates on the software
1694    cursor instead of the hardware cursor.
1695    
1696  If you want to disable the objectserver completely, you can execute  To this end, run "infocmp linux > linux-term", edit the file
1697  the command `chkconfig objectserver off' and reboot.  That may disable  `linux-term' to make both the "cnorm" and "cvvis" capabilities send
1698  some of the window system functionality, such as responding CDROM  the sequence "\E[?25h\E[?17;0;64c", and then run "tic linux-term" to
1699  icons.  produce a modified terminfo entry.
1700    
1701  You can also remove NIS support from the objectserver.  The SGI `admin'  Alternatively, if you want a blinking underscore as your Emacs cursor,
1702  FAQ has a detailed description on how to do that; see question 35  change the "cvvis" capability to send the "\E[?25h\E[?0c" command.
 ("Why isn't the objectserver working?").  The admin FAQ can be found at  
 ftp://viz.tamu.edu/pub/sgi/faq/.  
1703    
1704  * With certain fonts, when the cursor appears on a character, the  *** GNU/Linux: Error messages `internal facep []' happen on GNU/Linux systems.
 character doesn't appear--you get a solid box instead.  
1705    
1706  One user on a Linux-based GNU system reported that this problem went  There is a report that replacing libc.so.5.0.9 with libc.so.5.2.16
1707  away with installation of a new X server.  The failing server was  caused this to start happening.  People are not sure why, but the
1708  XFree86 3.1.1.  XFree86 3.1.2 works.  problem seems unlikely to be in Emacs itself.  Some suspect that it
1709    is actually Xlib which won't work with libc.so.5.2.16.
1710    
1711  * On SunOS 4.1.3, Emacs unpredictably crashes in _yp_dobind_soft.  Using the old library version is a workaround.
1712    
1713  This happens if you configure Emacs specifying just `sparc-sun-sunos4'  ** Mac OS X
 on a system that is version 4.1.3.  You must specify the precise  
 version number (or let configure figure out the configuration, which  
 it can do perfectly well for SunOS).  
1714    
1715  * On SunOS 4, Emacs processes keep going after you kill the X server  *** Mac OS X (Carbon): Environment Variables from dotfiles are ignored.
 (or log out, if you logged in using X).  
1716    
1717  Someone reported that recompiling with GCC 2.7.0 fixed this problem.  When starting Emacs from the Dock or the Finder on Mac OS X, the
1718    environment variables that are set up in dotfiles, such as .cshrc or
1719    .profile, are ignored.  This is because the Finder and Dock are not
1720    started from a shell, but instead from the Window Manager itself.
1721    
1722  * On AIX 4, some programs fail when run in a Shell buffer  The workaround for this is to create a .MacOSX/environment.plist file to
1723  with an error message like   No terminfo entry for "unknown".  setup these environment variables.  These environment variables will
1724    apply to all processes regardless of where they are started.
1725    For me information, see http://developer.apple.com/qa/qa2001/qa1067.html.
1726    
1727  On AIX, many terminal type definitions are not installed by default.  *** Mac OS X (Carbon): Process output truncated when using ptys.
 `unknown' is one of them.  Install the "Special Generic Terminal  
 Definitions" to make them defined.  
1728    
1729  * On SunOS, you get linker errors  There appears to be a problem with the implementation of pty's on the
1730     ld: Undefined symbol  Mac OS X that causes process output to be truncated.  To avoid this,
1731        _get_wmShellWidgetClass  leave process-connection-type set to its default value of nil.
       _get_applicationShellWidgetClass  
1732    
1733  The fix to this is to install patch 100573 for OpenWindows 3.0  ** FreeBSD
 or link libXmu statically.  
1734    
1735  * On AIX 4.1.2, linker error messages such as  *** FreeBSD 2.1.5: useless symbolic links remain in /tmp or other
1736    ld: 0711-212 SEVERE ERROR: Symbol .__quous, found in the global symbol table  directories that have the +t bit.
         of archive /usr/lib/libIM.a, was not defined in archive member shr.o.  
1737    
1738  This is a problem in libIM.a.  You can work around it by executing  This is because of a kernel bug in FreeBSD 2.1.5 (fixed in 2.2).
1739  these shell commands in the src subdirectory of the directory where  Emacs uses symbolic links to implement file locks.  In a directory
1740  you build Emacs:  with +t bit, the directory owner becomes the owner of the symbolic
1741    link, so that it cannot be removed by anyone else.
1742    
1743      cp /usr/lib/libIM.a .  If you don't like those useless links, you can let Emacs not to using
1744      chmod 664 libIM.a  file lock by adding #undef CLASH_DETECTION to config.h.
     ranlib libIM.a  
1745    
1746  Then change -lIM to ./libIM.a in the command to link temacs (in  *** FreeBSD: Getting a Meta key on the console.
 Makefile).  
1747    
1748  * Unpredictable segmentation faults on Solaris 2.3 and 2.4.  By default, neither Alt nor any other key acts as a Meta key on
1749    FreeBSD, but this can be changed using kbdcontrol(1).  Dump the
1750    current keymap to a file with the command
1751    
1752  A user reported that this happened in 19.29 when it was compiled with    $ kbdcontrol -d >emacs.kbd
 the Sun compiler, but not when he recompiled with GCC 2.7.0.  
1753    
1754  We do not know whether something in Emacs is partly to blame for this.  Edit emacs.kbd, and give the key you want to be the Meta key the
1755    definition `meta'.  For instance, if your keyboard has a ``Windows''
1756    key with scan code 105, change the line for scan code 105 in emacs.kbd
1757    to look like this
1758    
1759  * Emacs exits with "X protocol error" when run with an X server for    105   meta   meta   meta   meta   meta   meta   meta   meta    O
 MS-Windows.  
1760    
1761  A certain X server for Windows had a bug which caused this.  to make the Windows key the Meta key.  Load the new keymap with
 Supposedly the newer 32-bit version of this server doesn't have the  
 problem.  
1762    
1763  * Emacs crashes at startup on MSDOS.    $ kbdcontrol -l emacs.kbd
1764    
1765  Some users report that Emacs 19.29 requires dpmi memory management,  ** HP-UX
 and crashes on startup if the system does not have it.  We don't yet  
 know why this happens--perhaps these machines don't have enough real  
 memory, or perhaps something is wrong in Emacs or the compiler.  
 However, arranging to use dpmi support is a workaround.  
1766    
1767  You can find out if you have a dpmi host by running go32 without  *** HP/UX : Shell mode gives the message, "`tty`: Ambiguous".
 arguments; it will tell you if it uses dpmi memory.  For more  
 information about dpmi memory, consult the djgpp FAQ.  (djgpp  
 is the GNU C compiler as packaged for MSDOS.)  
1768    
1769  Compiling Emacs under MSDOS is extremely sensitive for proper memory  christos@theory.tn.cornell.edu says:
 configuration.  If you experience problems during compilation, consider  
 removing some or all memory resident programs (notably disk caches)  
 and make sure that your memory managers are properly configured.  See  
 the djgpp faq for configuration hints.  
1770    
1771  * A position you specified in .Xdefaults is ignored, using twm.  The problem is that in your .cshrc you have something that tries to
1772    execute `tty`. If you are not running the shell on a real tty then
1773    tty will print "not a tty". Csh expects one word in some places,
1774    but tty is giving it back 3.
1775    
1776  twm normally ignores "program-specified" positions.  The solution is to add a pair of quotes around `tty` to make it a single
1777  You can tell it to obey them with this command in your `.twmrc' file:  word:
1778    
1779    UsePPosition  "on"            #allow clients to request a position  if (`tty` == "/dev/console")
1780    
1781  * Compiling lib-src says there is no rule to make test-distrib.c.  should be changed to:
1782    
1783  This results from a bug in a VERY old version of GNU Sed.  To solve  if ("`tty`" == "/dev/console")
 the problem, install the current version of GNU Sed, then rerun  
 Emacs's configure script.  
1784    
1785  * Compiling wakeup, in lib-src, says it can't make wakeup.c.  Even better, move things that set up terminal sections out of .cshrc
1786    and into .login.
1787    
1788  This results from a bug in GNU Sed version 2.03.  To solve the  *** HP/UX: `Pid xxx killed due to text modification or page I/O error'.
 problem, install the current version of GNU Sed, then rerun Emacs's  
 configure script.  
1789    
1790  * On Sunos 4.1.1, there are errors compiling sysdep.c.  On HP/UX, you can get that error when the Emacs executable is on an NFS
1791    file system.  HP/UX responds this way if it tries to swap in a page and
1792    does not get a response from the server within a timeout whose default
1793    value is just ten seconds.
1794    
1795  If you get errors such as  If this happens to you, extend the timeout period.
1796    
1797      "sysdep.c", line 2017: undefined structure or union  *** HP/UX: Emacs is slow using X11R5.
     "sysdep.c", line 2017: undefined structure or union  
     "sysdep.c", line 2019: nodename undefined  
1798    
1799  This can result from defining LD_LIBRARY_PATH.  It is very tricky  This happens if you use the MIT versions of the X libraries--it
1800  to use that environment variable with Emacs.  The Emacs configure  doesn't run as fast as HP's version.  People sometimes use the version
1801  script links many test programs with the system libraries; you must  because they see the HP version doesn't have the libraries libXaw.a,
1802  make sure that the libraries available to configure are the same  libXmu.a, libXext.a and others.  HP/UX normally doesn't come with
1803  ones available when you build Emacs.  those libraries installed.  To get good performance, you need to
1804    install them and rebuild Emacs.
1805    
1806  * The right Alt key works wrong on German HP keyboards (and perhaps  *** HP/UX: The right Alt key works wrong on German HP keyboards (and perhaps
1807  other non-English HP keyboards too).  other non-English HP keyboards too).
1808    
1809  This is because HPUX defines the modifiers wrong in X.  Here is a  This is because HP-UX defines the modifiers wrong in X.  Here is a
1810  shell script to fix the problem; be sure that it is run after VUE  shell script to fix the problem; be sure that it is run after VUE
1811  configures the X server.  configures the X server.
1812    
# Line 2133  configures the X server. Line 1823  configures the X server.
1823      add mod2 = Mode_switch      add mod2 = Mode_switch
1824      EOF      EOF
1825    
1826  * The Emacs window disappears when you type M-q.  *** HP/UX: "Cannot find callback list" messages from dialog boxes in
1827    Emacs built with Motif.
1828    
1829  Some versions of the Open Look window manager interpret M-q as a quit  This problem resulted from a bug in GCC 2.4.5.  Newer GCC versions
1830  command for whatever window you are typing at.  If you want to use  such as 2.7.0 fix the problem.
 Emacs with that window manager, you should try to configure the window  
 manager to use some other command.   You can disable the  
 shortcut keys entirely by adding this line to ~/.OWdefaults:  
1831    
1832      OpenWindows.WindowMenuAccelerators: False  *** HP/UX: Emacs does not recognize the AltGr key.
1833    
1834  * Emacs does not notice when you release the mouse.  To fix this, set up a file ~/.dt/sessions/sessionetc with executable
1835    rights, containing this text:
1836    
1837  There are reports that this happened with (some) Microsoft mice and  --------------------------------
1838  that replacing the mouse made it stop.  xmodmap 2> /dev/null - << EOF
1839    keysym Alt_L = Meta_L
1840    keysym Alt_R = Meta_R
1841    EOF
1842    
1843  * Trouble using ptys on IRIX, or running out of ptys.  xmodmap - << EOF
1844    clear mod1
1845    keysym Mode_switch = NoSymbol
1846    add mod1 = Meta_L
1847    keysym Meta_R = Mode_switch
1848    add mod2 = Mode_switch
1849    EOF
1850    --------------------------------
1851    
1852  The program mkpts (which may be in `/usr/adm' or `/usr/sbin') needs to  *** HP/UX: Large file support is disabled.
 be set-UID to root, or non-root programs like Emacs will not be able  
 to allocate ptys reliably.  
1853    
1854  * On Irix 5.2, unexelfsgi.c can't find cmplrs/stsupport.h.  See the comments in src/s/hpux10.h.
1855    
1856  The file cmplrs/stsupport.h was included in the wrong file set in the  *** HP/UX 11.0: Emacs makes HP/UX 11.0 crash.
 Irix 5.2 distribution.  You can find it in the optional fileset  
 compiler_dev, or copy it from some other Irix 5.2 system.  A kludgy  
 workaround is to change unexelfsgi.c to include sym.h instead of  
 syms.h.  
1857    
1858  * Slow startup on Linux-based GNU systems.  This is a bug in HPUX; HPUX patch PHKL_16260 is said to fix it.
1859    
1860  People using systems based on the Linux kernel sometimes report that  ** AIX
 startup takes 10 to 15 seconds longer than `usual'.  
1861    
1862  This is because Emacs looks up the host name when it starts.  *** AIX: Trouble using ptys.
 Normally, this takes negligible time; the extra delay is due to  
 improper system configuration.  This problem can occur for both  
 networked and non-networked machines.  
1863    
1864  Here is how to fix the configuration.  It requires being root.  People often install the pty devices on AIX incorrectly.
1865    Use `smit pty' to reinstall them properly.
1866    
1867  ** Networked Case  *** AIXterm: Your Delete key sends a Backspace to the terminal.
1868    
1869  First, make sure the files `/etc/hosts' and `/etc/host.conf' both  The solution is to include in your .Xdefaults the lines:
 exist.  The first line in the `/etc/hosts' file should look like this  
 (replace HOSTNAME with your host name):  
1870    
1871      127.0.0.1      HOSTNAME     *aixterm.Translations: #override <Key>BackSpace: string(0x7f)
1872       aixterm*ttyModes: erase ^?
1873    
1874  Also make sure that the `/etc/host.conf' files contains the following  This makes your Backspace key send DEL (ASCII 127).
 lines:  
1875    
1876      order hosts, bind  *** AIX: You get this message when running Emacs:
     multi on  
1877    
1878  Any changes, permanent and temporary, to the host name should be      Could not load program emacs
1879  indicated in the `/etc/hosts' file, since it acts a limited local      Symbol smtcheckinit in csh is undefined
1880  database of addresses and names (e.g., some SLIP connections      Error was: Exec format error
 dynamically allocate ip addresses).  
1881    
1882  ** Non-Networked Case  or this one:
1883    
1884  The solution described in the networked case applies here as well.      Could not load program .emacs
1885  However, if you never intend to network your machine, you can use a      Symbol _system_con in csh is undefined
1886  simpler solution: create an empty `/etc/host.conf' file.  The command      Symbol _fp_trapsta in csh is undefined
1887  `touch /etc/host.conf' suffices to create the file.  The `/etc/hosts'      Error was: Exec format error
 file is not necessary with this approach.  
1888    
1889  * On Solaris 2.4, Dired hangs and C-g does not work.  Or Emacs hangs  These can happen when you try to run on AIX 3.2.5 a program that was
1890  forever waiting for termination of a subprocess that is a zombie.  compiled with 3.2.4.  The fix is to recompile.
1891    
1892  casper@fwi.uva.nl says the problem is in X11R6.  Rebuild libX11.so  *** AIX 3.2.4: Releasing Ctrl/Act key has no effect, if Shift is down.
 after changing the file xc/config/cf/sunLib.tmpl.  Change the lines  
1893    
1894      #if ThreadedX  Due to a feature of AIX, pressing or releasing the Ctrl/Act key is
1895      #define SharedX11Reqs -lthread  ignored when the Shift, Alt or AltGr keys are held down.  This can
1896      #endif  lead to the keyboard being "control-locked"--ordinary letters are
1897    treated as control characters.
1898    
1899  to:  You can get out of this "control-locked" state by pressing and
1900    releasing Ctrl/Act while not pressing or holding any other keys.
1901    
1902      #if OSMinorVersion < 4  *** AIX 4.2: Emacs gets a segmentation fault at startup.
     #if ThreadedX  
     #define SharedX11Reqs -lthread  
     #endif  
     #endif  
1903    
1904  Be sure also to edit x/config/cf/sun.cf so that OSMinorVersion is 4  If you are using IBM's xlc compiler, compile emacs.c
1905  (as it should be for Solaris 2.4).  The file has three definitions for  without optimization; that should avoid the problem.
 OSMinorVersion: the first is for x86, the second for SPARC under  
 Solaris, and the third for SunOS 4.  Make sure to update the  
 definition for your type of machine and system.  
1906    
1907  Then do `make Everything' in the top directory of X11R6, to rebuild  *** AIX: If linking fails because libXbsd isn't found, check if you
1908  the makefiles and rebuild X.  The X built this way work only on  are compiling with the system's `cc' and CFLAGS containing `-O5'.  If
1909  Solaris 2.4, not on 2.3.  so, you have hit a compiler bug.  Please make sure to re-configure
1910    Emacs so that it isn't compiled with `-O5'.
1911    
1912  For multithreaded X to work it is necessary to install patch  *** AIX 4.3.x or 4.4: Compiling fails.
 101925-02 to fix problems in header files [2.4].  You need  
 to reinstall gcc or re-run just-fixinc after installing that  
 patch.  
1913    
1914  However, Frank Rust <frust@iti.cs.tu-bs.de> used a simpler solution:  This could happen if you use /bin/c89 as your compiler, instead of
1915  he changed  the default `cc'.  /bin/c89 treats certain warnings, such as benign
1916      #define ThreadedX          YES  redefinitions of macros, as errors, and fails the build.  A solution
1917  to  is to use the default compiler `cc'.
     #define ThreadedX          NO  
 in sun.cf and did `make World' to rebuild X11R6.  Removing all  
 `-DXTHREAD*' flags and `-lthread' entries from lib/X11/Makefile and  
 typing 'make install' in that directory also seemed to work.  
1918    
1919  * With M-x enable-flow-control, you need to type C-\ twice  *** AIX 4: Some programs fail when run in a Shell buffer
1920    to do incremental search--a single C-\ gets no response.  with an error message like   No terminfo entry for "unknown".
1921    
1922  This has been traced to communicating with your machine via kermit,  On AIX, many terminal type definitions are not installed by default.
1923  with C-\ as the kermit escape character.  One solution is to use  `unknown' is one of them.  Install the "Special Generic Terminal
1924  another escape character in kermit.  One user did  Definitions" to make them defined.
1925    
1926     set escape-character 17  ** Solaris
1927    
1928  in his .kermrc file, to make C-q the kermit escape character.  We list bugs in current versions here.  Solaris 2.x and 4.x are covered in the
1929    section on legacy systems.
1930    
1931  * The Motif version of Emacs paints the screen a solid color.  *** On Solaris, C-x doesn't get through to Emacs when you use the console.
1932    
1933  This has been observed to result from the following X resource:  This is a Solaris feature (at least on Intel x86 cpus).  Type C-r
1934    C-r C-t, to toggle whether C-x gets through to Emacs.
1935    
1936     Emacs*default.attributeFont: -*-courier-medium-r-*-*-*-140-*-*-*-*-iso8859-*  *** Problem with remote X server on Suns.
1937    
1938  That the resource has this effect indicates a bug in something, but we  On a Sun, running Emacs on one machine with the X server on another
1939  do not yet know what.  If it is an Emacs bug, we hope someone can  may not work if you have used the unshared system libraries.  This
1940  explain what the bug is so we can fix it.  In the mean time, removing  is because the unshared libraries fail to use YP for host name lookup.
1941  the resource prevents the problem.  As a result, the host name you specify may not be recognized.
1942    
1943  * Emacs gets hung shortly after startup, on Sunos 4.1.3.  *** Emacs reports a BadAtom error (from X) running on Solaris 7 or 8.
1944    
1945  We think this is due to a bug in Sunos.  The word is that  This happens when Emacs was built on some other version of Solaris.
1946  one of these Sunos patches fixes the bug:  Rebuild it on Solaris 8.
1947    
1948  100075-11  100224-06  100347-03  100482-05  100557-02  100623-03  100804-03  101080-01  *** On Solaris, CTRL-t is ignored by Emacs when you use
1949  100103-12  100249-09             100496-02  100564-07  100630-02  100891-10  101134-01  the fr.ISO-8859-15 locale (and maybe other related locales).
 100170-09  100296-04  100377-09  100507-04  100567-04  100650-02  101070-01  101145-01  
 100173-10  100305-15  100383-06  100513-04  100570-05  100689-01  101071-03  101200-02  
 100178-09  100338-05  100421-03  100536-02  100584-05  100784-01  101072-01  101207-01  
1950    
1951  We don't know which of these patches really matter.  If you find out  You can fix this by editing the file:
 which ones, please inform bug-gnu-emacs@gnu.org.  
1952    
1953  * Emacs aborts while starting up, only when run without X.          /usr/openwin/lib/locale/iso8859-15/Compose
1954    
1955  This problem often results from compiling Emacs with GCC when GCC was  Near the bottom there is a line that reads:
 installed incorrectly.  The usual error in installing GCC is to  
 specify --includedir=/usr/include.  Installation of GCC makes  
 corrected copies of the system header files.  GCC is supposed to use  
 the corrected copies in preference to the original system headers.  
 Specifying --includedir=/usr/include causes the original system header  
 files to be used.  On some systems, the definition of ioctl in the  
 original system header files is invalid for ANSI C and causes Emacs  
 not to work.  
1956    
1957  The fix is to reinstall GCC, and this time do not specify --includedir          Ctrl<t> <quotedbl> <Y>                  : "\276"        threequarters
 when you configure it.  Then recompile Emacs.  Specifying --includedir  
 is appropriate only in very special cases and it should *never* be the  
 same directory where system header files are kept.  
1958    
1959  * On Solaris 2.x, GCC complains "64 bit integer types not supported"  that should read:
1960    
1961  This suggests that GCC is not installed correctly.  Most likely you          Ctrl<T> <quotedbl> <Y>                  : "\276"        threequarters
 are using GCC 2.7.2.3 (or earlier) on Solaris 2.6 (or later); this  
 does not work without patching.  To run GCC 2.7.2.3 on Solaris 2.6 or  
 later, you must patch fixinc.svr4 and reinstall GCC from scratch as  
 described in the Solaris FAQ  
 <http://www.wins.uva.nl/pub/solaris/solaris2.html>.  A better fix is  
 to upgrade to GCC 2.8.1 or later.  
1962    
1963  * The Compose key on a DEC keyboard does not work as Meta key.  Note the lower case <t>.  Changing this line should make C-t work.
1964    
1965  This shell command should fix it:  *** When using M-x dbx with the SparcWorks debugger, the `up' and `down'
1966    commands do not move the arrow in Emacs.
1967    
1968    xmodmap -e 'keycode 0xb1 = Meta_L'  You can fix this by adding the following line to `~/.dbxinit':
1969    
1970  * Regular expressions matching bugs on SCO systems.   dbxenv output_short_file_name off
1971    
1972  On SCO, there are problems in regexp matching when Emacs is compiled  ** Irix
 with the system compiler.  The compiler version is "Microsoft C  
 version 6", SCO 4.2.0h Dev Sys Maintenance Supplement 01/06/93; Quick  
 C Compiler Version 1.00.46 (Beta).  The solution is to compile with  
 GCC.  
1973    
1974  * On Sunos 4, you get the error ld: Undefined symbol __lib_version.  *** Irix 5.2: unexelfsgi.c can't find cmplrs/stsupport.h.
1975    
1976  This is the result of using cc or gcc with the shared library meant  The file cmplrs/stsupport.h was included in the wrong file set in the
1977  for acc (the Sunpro compiler).  Check your LD_LIBRARY_PATH and delete  Irix 5.2 distribution.  You can find it in the optional fileset
1978  /usr/lang/SC2.0.1 or some similar directory.  compiler_dev, or copy it from some other Irix 5.2 system.  A kludgy
1979    workaround is to change unexelfsgi.c to include sym.h instead of
1980    syms.h.
1981    
1982  * You can't select from submenus (in the X toolkit version).  *** Irix 5.3: "out of virtual swap space".
1983    
1984  On certain systems, mouse-tracking and selection in top-level menus  This message occurs when the system runs out of swap space due to too
1985  works properly with the X toolkit, but neither of them works when you  many large programs running.  The solution is either to provide more
1986  bring up a submenu (such as Bookmarks or Compare or Apply Patch, in  swap space or to reduce the number of large programs being run.  You
1987  the Files menu).  can check the current status of the swap space by executing the
1988    command `swap -l'.
1989    
1990  This works on most systems.  There is speculation that the failure is  You can increase swap space by changing the file /etc/fstab.  Adding a
1991  due to bugs in old versions of X toolkit libraries, but no one really  line like this:
1992  knows.  If someone debugs this and finds the precise cause, perhaps a  
1993  workaround can be found.  /usr/swap/swap.more     swap    swap    pri=3 0 0
1994    
1995    where /usr/swap/swap.more is a file previously created (for instance
1996    by using /etc/mkfile), will increase the swap space by the size of
1997    that file.  Execute `swap -m' or reboot the machine to activate the
1998    new swap area.  See the manpages for `swap' and `fstab' for further
1999    information.
2000    
2001    The objectserver daemon can use up lots of memory because it can be
2002    swamped with NIS information.  It collects information about all users
2003    on the network that can log on to the host.
2004    
2005    If you want to disable the objectserver completely, you can execute
2006    the command `chkconfig objectserver off' and reboot.  That may disable
2007    some of the window system functionality, such as responding CDROM
2008    icons.
2009    
2010    You can also remove NIS support from the objectserver.  The SGI `admin'
2011    FAQ has a detailed description on how to do that; see question 35
2012    ("Why isn't the objectserver working?").  The admin FAQ can be found at
2013    ftp://viz.tamu.edu/pub/sgi/faq/.
2014    
2015    *** Irix 5.3: Emacs crashes in utmpname.
2016    
2017  * Unusable default font on SCO 3.2v4.  This problem is fixed in Patch 3175 for Irix 5.3.
2018    It is also fixed in Irix versions 6.2 and up.
2019    
2020    *** Irix 6.0: Make tries (and fails) to build a program named unexelfsgi.
2021    
2022    A compiler bug inserts spaces into the string "unexelfsgi . o"
2023    in src/Makefile.  Edit src/Makefile, after configure is run,
2024    find that string, and take out the spaces.
2025    
2026    Compiler fixes in Irix 6.0.1 should eliminate this problem.
2027    
2028    *** Irix 6.5: Emacs crashes on the SGI R10K, when compiled with GCC.
2029    
2030    This seems to be fixed in GCC 2.95.
2031    
2032    *** Trouble using ptys on IRIX, or running out of ptys.
2033    
2034    The program mkpts (which may be in `/usr/adm' or `/usr/sbin') needs to
2035    be set-UID to root, or non-root programs like Emacs will not be able
2036    to allocate ptys reliably.
2037    
2038    ** SCO Unix and UnixWare
2039    
2040    *** SCO 3.2v4: Unusable default font.
2041    
2042  The Open Desktop environment comes with default X resource settings  The Open Desktop environment comes with default X resource settings
2043  that tell Emacs to use a variable-width font.  Emacs cannot use such  that tell Emacs to use a variable-width font.  Emacs cannot use such
# Line 2373  Open Desktop display. Line 2072  Open Desktop display.
2072  These resource files are not normally shared across a network of SCO  These resource files are not normally shared across a network of SCO
2073  machines; you must create the file on each machine individually.  machines; you must create the file on each machine individually.
2074    
2075  * rcs2log gives you the awk error message "too many fields".  *** Regular expressions matching bugs on SCO systems.
   
 This is due to an arbitrary limit in certain versions of awk.  
 The solution is to use gawk (GNU awk).  
   
 * Emacs is slow using X11R5 on HP/UX.  
   
 This happens if you use the MIT versions of the X libraries--it  
 doesn't run as fast as HP's version.  People sometimes use the version  
 because they see the HP version doesn't have the libraries libXaw.a,  
 libXmu.a, libXext.a and others.  HP/UX normally doesn't come with  
 those libraries installed.  To get good performance, you need to  
 install them and rebuild Emacs.  
2076    
2077  * Loading fonts is very slow.  On SCO, there are problems in regexp matching when Emacs is compiled
2078    with the system compiler.  The compiler version is "Microsoft C
2079    version 6", SCO 4.2.0h Dev Sys Maintenance Supplement 01/06/93; Quick
2080    C Compiler Version 1.00.46 (Beta).  The solution is to compile with
2081    GCC.
2082    
2083  You might be getting scalable fonts instead of precomputed bitmaps.  *** UnixWare 2.1: Error 12 (virtual memory exceeded) when dumping Emacs.
 Known scalable font directories are "Type1" and "Speedo".  A font  
 directory contains scalable fonts if it contains the file  
 "fonts.scale".  
2084    
2085  If this is so, re-order your X windows font path to put the scalable  Paul Abrahams (abrahams@acm.org) reports that with the installed
2086  font directories last.  See the documentation of `xset' for details.  virtual memory settings for UnixWare 2.1.2, an Error 12 occurs during
2087    the "make" that builds Emacs, when running temacs to dump emacs.  That
2088    error indicates that the per-process virtual memory limit has been
2089    exceeded.  The default limit is probably 32MB.  Raising the virtual
2090    memory limit to 40MB should make it possible to finish building Emacs.
2091    
2092  With some X servers, it may be necessary to take the scalable font  You can do this with the command `ulimit' (sh) or `limit' (csh).
2093  directories out of your path entirely, at least for Emacs 19.26.  But you have to be root to do it.
 Changes in the future may make this unnecessary.  
2094    
2095  * On AIX 3.2.4, releasing Ctrl/Act key has no effect, if Shift is down.  According to Martin Sohnius, you can also retune this in the kernel:
2096    
2097  Due to a feature of AIX, pressing or releasing the Ctrl/Act key is      # /etc/conf/bin/idtune SDATLIM 33554432         ## soft data size limit
2098  ignored when the Shift, Alt or AltGr keys are held down.  This can      # /etc/conf/bin/idtune HDATLIM 33554432         ## hard "
2099  lead to the keyboard being "control-locked"--ordinary letters are      # /etc/conf/bin/idtune SVMMSIZE unlimited       ## soft process size limit
2100  treated as control characters.      # /etc/conf/bin/idtune HVMMSIZE unlimited       ## hard "
2101        # /etc/conf/bin/idbuild -B
2102    
2103  You can get out of this "control-locked" state by pressing and  (He recommends you not change the stack limit, though.)
2104  releasing Ctrl/Act while not pressing or holding any other keys.  These changes take effect when you reboot.
2105    
2106  * display-time causes kernel problems on ISC systems.  * Runtime problems specific to MS-Windows
2107    
2108  Under Interactive Unix versions 3.0.1 and 4.0 (and probably other  ** Emacs exits with "X protocol error" when run with an X server for MS-Windows.
 versions), display-time causes the loss of large numbers of STREVENT  
 cells.  Eventually the kernel's supply of these cells is exhausted.  
 This makes emacs and the whole system run slow, and can make other  
 processes die, in particular pcnfsd.  
2109    
2110  Other emacs functions that communicate with remote processes may have  A certain X server for Windows had a bug which caused this.
2111  the same problem.  Display-time seems to be far the worst.  Supposedly the newer 32-bit version of this server doesn't have the
2112    problem.
2113    
2114  The only known fix: Don't run display-time.  ** Known problems with the MS-Windows port of Emacs 21.2.
2115    
2116  * On Solaris, C-x doesn't get through to Emacs when you use the console.  Frames are not refreshed while the File or Font dialog or a pop-up menu
2117    is displayed. This also means help text for pop-up menus is not
2118    displayed at all.  This is because message handling under Windows is
2119    synchronous, so we cannot handle repaint (or any other) messages while
2120    waiting for a system function to return the result of the dialog or
2121    pop-up menu interaction.
2122    
2123  This is a Solaris feature (at least on Intel x86 cpus).  Type C-r  Windows 95 and Windows NT up to version 4.0 do not support help text
2124  C-r C-t, to toggle whether C-x gets through to Emacs.  for menus.  Help text is only available in later versions of Windows.
2125    
2126  * Error message `Symbol's value as variable is void: x', followed by  There are problems with display if mouse-tracking is enabled and the
2127    segmentation fault and core dump.  mouse is moved off a frame, over another frame then back over the first
2128    frame.  A workaround is to click the left mouse button inside the frame
2129    after moving back into it.
2130    
2131  This has been tracked to a bug in tar!  People report that tar erroneously  Some minor flickering still persists during mouse-tracking, although
2132  added a line like this at the beginning of files of Lisp code:  not as severely as in 21.1.
2133    
2134     x FILENAME, N bytes, B tape blocks  Emacs can sometimes abort when non-ASCII text, possibly with null
2135    characters, is copied and pasted into a buffer.
2136    
2137  If your tar has this problem, install GNU tar--if you can manage to  An inactive cursor remains in an active window after the Windows
2138  untar it :-).  Manager driven switch of the focus, until a key is pressed.
2139    
2140  * Link failure when using acc on a Sun.  Windows input methods are not recognized by Emacs (as of v21.2).  Some
2141    of these input methods cause the keyboard to send characters encoded
2142    in the appropriate coding system (e.g., ISO 8859-1 for Latin-1
2143    characters, ISO 8859-8 for Hebrew characters, etc.).  To make this
2144    work, set the keyboard coding system to the appropriate value after
2145    you activate the Windows input method.  For example, if you activate
2146    the Hebrew input method, type "C-x RET k iso-8859-8 RET".  (Emacs
2147    ought to recognize the Windows language-change event and set up the
2148    appropriate keyboard encoding automatically, but it doesn't do that
2149    yet.)
2150    
2151  To use acc, you need additional options just before the libraries, such as  The %b specifier for format-time-string does not produce abbreviated
2152    month names with consistent widths for some locales on some versions
2153    of Windows. This is caused by a deficiency in the underlying system
2154    library function.
2155    
2156     /usr/lang/SC2.0.1/values-Xt.o -L/usr/lang/SC2.0.1/cg87 -L/usr/lang/SC2.0.1  ** Problems running Perl under Emacs on MS-Windows NT/95.
2157    
2158  and you need to add -lansi just before -lc.  `perl -de 0' just hangs when executed in an Emacs subshell.
2159    The fault lies with Perl (indirectly with Windows NT/95).
2160    
2161  The precise file names depend on the compiler version, so we  The problem is that the Perl debugger explicitly opens a connection to
2162  cannot easily arrange to supply them.  "CON", which is the DOS/NT equivalent of "/dev/tty", for interacting
2163    with the user.
2164    
2165  * Link failure on IBM AIX 1.3 ptf 0013.  On Unix, this is okay, because Emacs (or the shell?) creates a
2166    pseudo-tty so that /dev/tty is really the pipe Emacs is using to
2167    communicate with the subprocess.
2168    
2169  There is a real duplicate definition of the function `_slibc_free' in  On NT, this fails because CON always refers to the handle for the
2170  the library /lib/libc_s.a (just do nm on it to verify).  The  relevant console (approximately equivalent to a tty), and cannot be
2171  workaround/fix is:  redirected to refer to the pipe Emacs assigned to the subprocess as
2172    stdin.
2173    
2174      cd /lib  A workaround is to modify perldb.pl to use STDIN/STDOUT instead of CON.
     ar xv libc_s.a NLtmtime.o  
     ar dv libc_s.a NLtmtime.o  
2175    
2176  * Undefined symbols _dlopen, _dlsym and/or _dlclose on a Sun.  For Perl 4:
2177    
2178  If you see undefined symbols _dlopen, _dlsym, or _dlclose when linking      *** PERL/LIB/PERLDB.PL.orig Wed May 26 08:24:18 1993
2179  with -lX11, compile and link against the file mit/util/misc/dlsym.c in      --- PERL/LIB/PERLDB.PL      Mon Jul 01 15:28:16 1996
2180  the MIT X11R5 distribution.  Alternatively, link temacs using shared      ***************
2181  libraries with s/sunos4shr.h.  (This doesn't work if you use the X      *** 68,74 ****
2182  toolkit.)            $rcfile=".perldb";
2183          }
2184          else {
2185        !     $console = "con";
2186              $rcfile="perldb.ini";
2187          }
2188    
2189  If you get the additional error that the linker could not find      --- 68,74 ----
2190  lib_version.o, try extracting it from X11/usr/lib/X11/libvim.a in            $rcfile=".perldb";
2191  X11R4, then use it in the link.        }
2192          else {
2193        !     $console = "";
2194              $rcfile="perldb.ini";
2195          }
2196    
 * Error messages `Wrong number of arguments: #<subr where-is-internal>, 5'  
2197    
2198  This typically results from having the powerkey library loaded.      For Perl 5:
2199  Powerkey was designed for Emacs 19.22.  It is obsolete now because      *** perl/5.001/lib/perl5db.pl.orig  Sun Jun 04 21:13:40 1995
2200  Emacs 19 now has this feature built in; and powerkey also calls      --- perl/5.001/lib/perl5db.pl       Mon Jul 01 17:00:08 1996
2201  where-is-internal in an obsolete way.      ***************
2202        *** 22,28 ****
2203              $rcfile=".perldb";
2204          }
2205          elsif (-e "con") {
2206        !     $console = "con";
2207              $rcfile="perldb.ini";
2208          }
2209          else {
2210        --- 22,28 ----
2211              $rcfile=".perldb";
2212          }
2213          elsif (-e "con") {
2214        !     $console = "";
2215              $rcfile="perldb.ini";
2216          }
2217          else {
2218    
2219  So the fix is to arrange not to load powerkey.  ** On MS-Windows 95, Alt-f6 does not get through to Emacs.
2220    
2221  * In Shell mode, you get a ^M at the end of every line.  This character seems to be trapped by the kernel in Windows 95.
2222    You can enter M-f6 by typing ESC f6.
2223    
2224  This happens to people who use tcsh, because it is trying to be too  ** Typing Alt-Shift has strange effects on MS-Windows.
 smart.  It sees that the Shell uses terminal type `unknown' and turns  
 on the flag to output ^M at the end of each line.  You can fix the  
 problem by adding this to your .cshrc file:  
2225    
2226      if ($?EMACS) then  This combination of keys is a command to change keyboard layout.  If
2227          if ($EMACS == "t") then  you proceed to type another non-modifier key before you let go of Alt
2228              unset edit  and Shift, the Alt and Shift act as modifiers in the usual way.  A
2229              stty  -icrnl -onlcr -echo susp ^Z  more permanent work around is to change it to another key combination,
2230          endif  or disable it in the keyboard control panel.
     endif  
2231    
2232  * An error message such as `X protocol error: BadMatch (invalid  ** Interrupting Cygwin port of Bash from Emacs doesn't work.
 parameter attributes) on protocol request 93'.  
2233    
2234  This comes from having an invalid X resource, such as  Cygwin 1.x builds of the ported Bash cannot be interrupted from the
2235     emacs*Cursor:   black  MS-Windows version of Emacs.  This is due to some change in the Bash
2236  (which is invalid because it specifies a color name for something  port or in the Cygwin library which apparently make Bash ignore the
2237  that isn't a color.)  keyboard interrupt event sent by Emacs to Bash.  (Older Cygwin ports
2238    of Bash, up to b20.1, did receive SIGINT from Emacs.)
2239    
2240  The fix is to correct your X resources.  ** Accessing remote files with ange-ftp hangs the MS-Windows version of Emacs.
2241    
2242  * Undefined symbols when linking on Sunos 4.1 using --with-x-toolkit.  If the FTP client is the Cygwin port of GNU `ftp', this appears to be
2243    due to some bug in the Cygwin DLL or some incompatibility between it
2244    and the implementation of asynchronous subprocesses in the Windows
2245    port of Emacs.  Specifically, some parts of the FTP server responses
2246    are not flushed out, apparently due to buffering issues, which
2247    confuses ange-ftp.
2248    
2249  If you get the undefined symbols _atowc _wcslen, _iswprint, _iswspace,  The solution is to downgrade to an older version of the Cygwin DLL
2250  _iswcntrl, _wcscpy, and _wcsncpy, then you need to add -lXwchar after  (version 1.3.2 was reported to solve the problem), or use the stock
2251  -lXaw in the command that links temacs.  Windows FTP client, usually found in the `C:\WINDOWS' or 'C:\WINNT'
2252    directory.  To force ange-ftp use the stock Windows client, set the
2253    variable `ange-ftp-ftp-program-name' to the absolute file name of the
2254    client's executable.  For example:
2255    
2256  This problem seems to arise only when the international language   (setq ange-ftp-ftp-program-name "c:/windows/ftp.exe")
 extensions to X11R5 are installed.  
2257    
2258  * Typing C-c C-c in Shell mode kills your X server.  If you want to stick with the Cygwin FTP client, you can work around
2259    this problem by putting this in your `.emacs' file:
2260    
2261  This happens with Linux kernel 1.0 thru 1.04, approximately.  The workaround is   (setq ange-ftp-ftp-program-args '("-i" "-n" "-g" "-v" "--prompt" "")
 to define SIGNALS_VIA_CHARACTERS in config.h and recompile Emacs.  
 Newer Linux kernel versions don't have this problem.  
2262    
2263  * src/Makefile and lib-src/Makefile are truncated--most of the file missing.  ** lpr commands don't work on MS-Windows with some cheap printers.
2264    
2265  This can happen if configure uses GNU sed version 2.03.  That version  This problem may also strike other platforms, but the solution is
2266  had a bug.  GNU sed version 2.05 works properly.  likely to be a global one, and not Emacs specific.
2267    
2268  * Slow startup on X11R6 with X windows.  Many cheap inkjet, and even some cheap laser printers, do not
2269    print plain text anymore, they will only print through graphical
2270    printer drivers. A workaround on MS-Windows is to use Windows' basic
2271    built in editor to print (this is possibly the only useful purpose it
2272    has):
2273    
2274  If Emacs takes two minutes to start up on X11R6, see if your X  (setq printer-name "")         ;; notepad takes the default
2275  resources specify any Adobe fonts.  That causes the type-1 font  (setq lpr-command "notepad")   ;; notepad
2276  renderer to start up, even if the font you asked for is not a type-1  (setq lpr-switches nil)        ;; not needed
2277  font.  (setq lpr-printer-switch "/P") ;; run notepad as batch printer
2278    
2279  One way to avoid this problem is to eliminate the type-1 fonts from  ** Antivirus software interacts badly with the MS-Windows version of Emacs.
 your font path, like this:  
2280    
2281          xset -fp /usr/X11R6/lib/X11/fonts/Type1/  The usual manifestation of these problems is that subprocesses don't
2282    work or even wedge the entire system.  In particular, "M-x shell RET"
2283    was reported to fail to work.  But other commands also sometimes don't
2284    work when an antivirus package is installed.
2285    
2286  * Pull-down menus appear in the wrong place, in the toolkit version of Emacs.  The solution is to switch the antivirus software to a less aggressive
2287    mode (e.g., disable the ``auto-protect'' feature), or even uninstall
2288    or disable it entirely.
2289    
2290  An X resource of this form can cause the problem:  ** On MS-Windows 95/98/ME, subprocesses do not terminate properly.
2291    
2292     Emacs*geometry:      80x55+0+0  This is a limitation of the Operating System, and can cause problems
2293    when shutting down Windows. Ensure that all subprocesses are exited
2294    cleanly before exiting Emacs. For more details, see the FAQ at
2295    http://www.gnu.org/software/emacs/windows/.
2296    
2297  This resource is supposed to apply, and does apply, to the menus  ** MS-Windows 95/98/ME crashes when Emacs invokes non-existent programs.
 individually as well as to Emacs frames.  If that is not what you  
 want, rewrite the resource.  
2298    
2299  To check thoroughly for such resource specifications, use `xrdb  When a program you are trying to run is not found on the PATH,
2300  -query' to see what resources the X server records, and also look at  Windows might respond by crashing or locking up your system.  In
2301  the user's ~/.Xdefaults and ~/.Xdefaults-* files.  particular, this has been reported when trying to compile a Java
2302    program in JDEE when javac.exe is installed, but not on the system
2303    PATH.
2304    
2305  * --with-x-toolkit version crashes when used with shared libraries.  ** Pressing the mouse button on MS-Windows does not give a mouse-2 event.
2306    
2307  On some systems, including Sunos 4 and DGUX 5.4.2 and perhaps others,  This is usually a problem with the mouse driver. Because most Windows
2308  unexec doesn't work properly with the shared library for the X  programs do not do anything useful with the middle mouse button, many
2309  toolkit.  You might be able to work around this by using a nonshared  mouse drivers allow you to define the wheel press to do something
2310  libXt.a library.  The real fix is to upgrade the various versions of  different. Some drivers do not even have the option to generate a
2311  unexec and/or ralloc.  We think this has been fixed on Sunos 4  middle button press. In such cases, setting the wheel press to
2312  and Solaris in version 19.29.  "scroll" sometimes works if you press the button twice. Trying a
2313    generic mouse driver might help.
2314    
2315  * `make install' fails on install-doc with `Error 141'.  ** Scrolling the mouse wheel on MS-Windows always scrolls the top window.
2316    
2317  This happens on Ultrix 4.2 due to failure of a pipeline of tar  This is another common problem with mouse drivers. Instead of
2318  commands.  We don't know why they fail, but the bug seems not to be in  generating scroll events, some mouse drivers try to fake scroll bar
2319  Emacs.  The workaround is to run the shell command in install-doc by  movement. But they are not intelligent enough to handle multiple
2320  hand.  scroll bars within a frame. Trying a generic mouse driver might help.
2321    
2322  * --with-x-toolkit option configures wrong on BSD/386.  ** Mail sent through Microsoft Exchange in some encodings appears to be
2323    mangled and is not seen correctly in Rmail or Gnus.  We don't know
2324    exactly what happens, but it isn't an Emacs problem in cases we've
2325    seen.
2326    
2327  This problem is due to bugs in the shell in version 1.0 of BSD/386.  ** On MS-Windows, you cannot use the right-hand ALT key and the left-hand
2328  The workaround is to edit the configure file to use some other shell,  CTRL key together to type a Control-Meta character.
 such as bash.  
2329    
2330  * Subprocesses remain, hanging but not zombies, on Sunos 5.3.  This is a consequence of a misfeature beyond Emacs's control.
2331    
2332  A bug in Sunos 5.3 causes Emacs subprocesses to remain after Emacs  Under Windows, the AltGr key on international keyboards generates key
2333  exits.  Sun patch # 101415-02 is part of the fix for this, but it only  events with the modifiers Right-Alt and Left-Ctrl.  Since Emacs cannot
2334  applies to ptys, and doesn't fix the problem with subprocesses  distinguish AltGr from an explicit Right-Alt and Left-Ctrl
2335  communicating through pipes.  combination, whenever it sees Right-Alt and Left-Ctrl it assumes that
2336    AltGr has been pressed.  The variable `w32-recognize-altgr' can be set
2337    to nil to tell Emacs that AltGr is really Ctrl and Alt.
2338    
2339  * Mail is lost when sent to local aliases.  ** Under some X-servers running on MS-Windows, Emacs' display is incorrect.
2340    
2341  Many emacs mail user agents (VM and rmail, for instance) use the  The symptoms are that Emacs does not completely erase blank areas of the
2342  sendmail.el library.  This library can arrange for mail to be  screen during scrolling or some other screen operations (e.g., selective
2343  delivered by passing messages to the /usr/lib/sendmail (usually)  display or when killing a region).  M-x recenter will cause the screen
2344  program .  In doing so, it passes the '-t' flag to sendmail, which  to be completely redisplayed and the "extra" characters will disappear.
 means that the name of the recipient of the message is not on the  
 command line and, therefore, that sendmail must parse the message to  
 obtain the destination address.  
2345    
2346  There is a bug in the SunOS4.1.1 and SunOS4.1.3 versions of sendmail.  This is known to occur under Exceed 6, and possibly earlier versions
2347  In short, when given the -t flag, the SunOS sendmail won't recognize  as well; it is reportedly solved in version 6.2.0.16 and later.  The
2348  non-local (i.e. NIS) aliases.  It has been reported that the Solaris  problem lies in the X-server settings.
 2.x versions of sendmail do not have this bug.  For those using SunOS  
 4.1, the best fix is to install sendmail V8 or IDA sendmail (which  
 have other advantages over the regular sendmail as well).  At the time  
 of this writing, these official versions are available:  
2349    
2350   Sendmail V8 on ftp.cs.berkeley.edu in /ucb/sendmail:  There are reports that you can solve the problem with Exceed by
2351     sendmail.8.6.9.base.tar.Z (the base system source & documentation)  running `Xconfig' from within NT, choosing "X selection", then
2352     sendmail.8.6.9.cf.tar.Z   (configuration files)  un-checking the boxes "auto-copy X selection" and "auto-paste to X
2353     sendmail.8.6.9.misc.tar.Z (miscellaneous support programs)  selection".
    sendmail.8.6.9.xdoc.tar.Z (extended documentation, with postscript)  
2354    
2355   IDA sendmail on vixen.cso.uiuc.edu in /pub:  Of this does not work, please inform bug-gnu-emacs@gnu.org.  Then
2356     sendmail-5.67b+IDA-1.5.tar.gz  please call support for your X-server and see if you can get a fix.
2357    If you do, please send it to bug-gnu-emacs@gnu.org so we can list it
2358    here.
2359    
2360  * On AIX, you get this message when running Emacs:  * Build-time problems
2361    
2362      Could not load program emacs  ** Configuration
     Symbol smtcheckinit in csh is undefined  
     Error was: Exec format error  
2363    
2364  or this one:  *** The `configure' script doesn't find the jpeg library.
2365    
2366      Could not load program .emacs  There are reports that this happens on some systems because the linker
2367      Symbol _system_con in csh is undefined  by default only looks for shared libraries, but jpeg distribution by
2368      Symbol _fp_trapsta in csh is undefined  default only installs a nonshared version of the library, `libjpeg.a'.
     Error was: Exec format error  
2369    
2370  These can happen when you try to run on AIX 3.2.5 a program that was  If this is the problem, you can configure the jpeg library with the
2371  compiled with 3.2.4.  The fix is to recompile.  `--enable-shared' option and then rebuild libjpeg.  This produces a
2372    shared version of libjpeg, which you need to install.  Finally, rerun
2373    the Emacs configure script, which should now find the jpeg library.
2374    Alternatively, modify the generated src/Makefile to link the .a file
2375    explicitly, and edit src/config.h to define HAVE_JPEG.
2376    
2377  * On AIX, you get this compiler error message:  *** AIX: You get this compiler error message:
2378    
2379      Processing include file ./XMenuInt.h      Processing include file ./XMenuInt.h
2380          1501-106: (S) Include file X11/Xlib.h not found.          1501-106: (S) Include file X11/Xlib.h not found.
# Line 2633  This means your system was installed wit Line 2383  This means your system was installed wit
2383  libraries.  You have to find your sipo (bootable tape) and install  libraries.  You have to find your sipo (bootable tape) and install
2384  X11Dev... with smit.  X11Dev... with smit.
2385    
2386  * You "lose characters" after typing Compose Character key.  ** Compilation
   
 This is because the Compose Character key is defined as the keysym  
 Multi_key, and Emacs (seeing that) does the proper X11  
 character-composition processing.  If you don't want your Compose key  
 to do that, you can redefine it with xmodmap.  
   
 For example, here's one way to turn it into a Meta key:  
   
     xmodmap -e "keysym Multi_key = Meta_L"  
   
 If all users at your site of a particular keyboard prefer Meta to  
 Compose, you can make the remapping happen automatically by adding the  
 xmodmap command to the xdm setup script for that display.  
   
 * C-z just refreshes the screen instead of suspending Emacs.  
   
 You are probably using a shell that doesn't support job control, even  
 though the system itself is capable of it.  Either use a different shell,  
 or set the variable `cannot-suspend' to a non-nil value.  
   
 * Watch out for .emacs files and EMACSLOADPATH environment vars  
   
 These control the actions of Emacs.  
 ~/.emacs is your Emacs init file.  
 EMACSLOADPATH overrides which directories the function  
 "load" will search.  
   
 If you observe strange problems, check for these and get rid  
 of them, then try again.  
   
 * After running emacs once, subsequent invocations crash.  
   
 Some versions of SVR4 have a serious bug in the implementation of the  
 mmap () system call in the kernel; this causes emacs to run correctly  
 the first time, and then crash when run a second time.  
   
 Contact your vendor and ask for the mmap bug fix; in the mean time,  
 you may be able to work around the problem by adding a line to your  
 operating system description file (whose name is reported by the  
 configure script) that reads:  
 #define SYSTEM_MALLOC  
 This makes Emacs use memory less efficiently, but seems to work around  
 the kernel bug.  
   
 * Inability to send an Alt-modified key, when Emacs is communicating  
 directly with an X server.  
   
 If you have tried to bind an Alt-modified key as a command, and it  
 does not work to type the command, the first thing you should check is  
 whether the key is getting through to Emacs.  To do this, type C-h c  
 followed by the Alt-modified key.  C-h c should say what kind of event  
 it read.  If it says it read an Alt-modified key, then make sure you  
 have made the key binding correctly.  
   
 If C-h c reports an event that doesn't have the Alt modifier, it may  
 be because your X server has no key for the Alt modifier.  The X  
 server that comes from MIT does not set up the Alt modifier by  
 default.  
2387    
2388  If your keyboard has keys named Alt, you can enable them as follows:  *** Building Emacs over NFS fails with ``Text file busy''.
   
     xmodmap -e 'add mod2 = Alt_L'  
     xmodmap -e 'add mod2 = Alt_R'  
   
 If the keyboard has just one key named Alt, then only one of those  
 commands is needed.  The modifier `mod2' is a reasonable choice if you  
 are using an unmodified MIT version of X.  Otherwise, choose any  
 modifier bit not otherwise used.  
2389    
2390  If your keyboard does not have keys named Alt, you can use some other  This was reported to happen when building Emacs on a GNU/Linux system
2391  keys.  Use the keysym command in xmodmap to turn a function key (or  (RedHat Linux 6.2) using a build directory automounted from Solaris
2392  some other 'spare' key) into Alt_L or into Alt_R, and then use the  (SunOS 5.6) file server, but it might not be limited to that
2393  commands show above to make them modifier keys.  configuration alone.  Presumably, the NFS server doesn't commit the
2394    files' data to disk quickly enough, and the Emacs executable file is
2395  Note that if you have Alt keys but no Meta keys, Emacs translates Alt  left ``busy'' for several seconds after Emacs has finished dumping
2396  into Meta.  This is because of the great importance of Meta in Emacs.  itself.  This causes the subsequent commands which invoke the dumped
2397    Emacs executable to fail with the above message.
 * `Pid xxx killed due to text modification or page I/O error'  
   
 On HP/UX, you can get that error when the Emacs executable is on an NFS  
 file system.  HP/UX responds this way if it tries to swap in a page and  
 does not get a response from the server within a timeout whose default  
 value is just ten seconds.  
   
 If this happens to you, extend the timeout period.  
2398    
2399  * `expand-file-name' fails to work on any but the machine you dumped Emacs on.  In some of these cases, a time skew between the NFS server and the
2400    machine where Emacs is built is detected and reported by GNU Make
2401    (it says that some of the files have modification time in the future).
2402    This might be a symptom of NFS-related problems.
2403    
2404  On Ultrix, if you use any of the functions which look up information  If the NFS server runs on Solaris, apply the Solaris patch 105379-05
2405  in the passwd database before dumping Emacs (say, by using  (Sunos 5.6: /kernel/misc/nfssrv patch).  If that doesn't work, or if
2406  expand-file-name in site-init.el), then those functions will not work  you have a different version of the OS or the NFS server, you can
2407  in the dumped Emacs on any host but the one Emacs was dumped on.  force the NFS server to use 1KB blocks, which was reported to fix the
2408    problem albeit at a price of slowing down file I/O.  You can force 1KB
2409    blocks by specifying the "-o  rsize=1024,wsize=1024" options to the
2410    `mount' command, or by adding ",rsize=1024,wsize=1024" to the mount
2411    options in the appropriate system configuration file, such as
2412    `/etc/auto.home'.
2413    
2414  The solution?  Don't use expand-file-name in site-init.el, or in  Alternatively, when Make fails due to this problem, you could wait for
2415  anything it loads.  Yuck - some solution.  a few seconds and then invoke Make again.  In one particular case,
2416    waiting for 10 or more seconds between the two Make invocations seemed
2417    to work around the problem.
2418    
2419  I'm not sure why this happens; if you can find out exactly what is  Similar problems can happen if your machine NFS-mounts a directory
2420  going on, and perhaps find a fix or a workaround, please let us know.  onto itself.  Suppose the Emacs sources live in `/usr/local/src' and
2421  Perhaps the YP functions cache some information, the cache is included  you are working on the host called `marvin'.  Then an entry in the
2422  in the dumped Emacs, and is then inaccurate on any other host.  `/etc/fstab' file like the following is asking for trouble:
2423    
2424  * On some variants of SVR4, Emacs does not work at all with X.      marvin:/usr/local/src /usr/local/src ...options.omitted...
2425    
2426  Try defining BROKEN_FIONREAD in your config.h file.  If this solves  The solution is to remove this line from `etc/fstab'.
 the problem, please send a bug report to tell us this is needed; be  
 sure to say exactly what type of machine and system you are using.  
2427    
2428  * Linking says that the functions insque and remque are undefined.  *** Building Emacs with GCC 2.9x fails in the `src' directory.
2429    
2430  Change oldXMenu/Makefile by adding insque.o to the variable OBJS.  This may happen if you use a development version of GNU `cpp' from one
2431    of the GCC snapshots between Oct 2000 and Feb 2001, or from a released
2432    version of GCC newer than 2.95.2 which was prepared around those
2433    dates; similar problems were reported with some snapshots of GCC 3.1
2434    around Sep 30 2001.  The preprocessor in those versions is
2435    incompatible with a traditional Unix cpp (e.g., it expands ".." into
2436    ". .", which breaks relative file names that reference the parent
2437    directory; or inserts TAB characters before lines that set Make
2438    variables).
2439    
2440  * Emacs fails to understand most Internet host names, even though  The solution is to make sure the preprocessor is run with the
2441  the names work properly with other programs on the same system.  `-traditional' option.  The `configure' script does that automatically
2442  * Emacs won't work with X-windows if the value of DISPLAY is HOSTNAME:0.  when it detects the known problems in your cpp, but you might hit some
2443  * GNUs can't make contact with the specified host for nntp.  unknown ones.  To force the `configure' script to use `-traditional',
2444    run the script like this:
2445    
2446  This typically happens on Suns and other systems that use shared    CPP='gcc -E -traditional' ./configure ...
 libraries.  The cause is that the site has installed a version of the  
 shared library which uses a name server--but has not installed a  
 similar version of the unshared library which Emacs uses.  
2447    
2448  The result is that most programs, using the shared library, work with  (replace the ellipsis "..." with any additional arguments you pass to
2449  the nameserver, but Emacs does not.  the script).
2450    
2451  The fix is to install an unshared library that corresponds to what you  Note that this problem does not pertain to the MS-Windows port of
2452  installed in the shared library, and then relink Emacs.  Emacs, since it doesn't use the preprocessor to generate Makefiles.
2453    
2454  On SunOS 4.1, simply define HAVE_RES_INIT.  *** src/Makefile and lib-src/Makefile are truncated--most of the file missing.
2455    *** Compiling wakeup, in lib-src, says it can't make wakeup.c.
2456    
2457  If you have already installed the name resolver in the file libresolv.a,  This can happen if configure uses GNU sed version 2.03.  That version
2458  then you need to compile Emacs to use that library.  The easiest way to  had a bug.  GNU sed version 2.05 works properly.To solve the
2459  do this is to add to config.h a definition of LIBS_SYSTEM, LIBS_MACHINE  problem, install the current version of GNU Sed, then rerun Emacs's
2460  or LIB_STANDARD which uses -lresolv.  Watch out!  If you redefine a macro  configure script.
 that is already in use in your configuration to supply some other libraries,  
 be careful not to lose the others.  
2461    
2462  Thus, you could start by adding this to config.h:  *** Compiling lib-src says there is no rule to make test-distrib.c.
2463    
2464  #define LIBS_SYSTEM -lresolv  This results from a bug in a VERY old version of GNU Sed.  To solve
2465    the problem, install the current version of GNU Sed, then rerun
2466    Emacs's configure script.
2467    
2468  Then if this gives you an error for redefining a macro, and you see that  *** Building the MS-Windows port with Cygwin GCC can fail.
 the s- file defines LIBS_SYSTEM as -lfoo -lbar, you could change config.h  
 again to say this:  
2469    
2470  #define LIBS_SYSTEM -lresolv -lfoo -lbar  Emacs may not build using recent Cygwin builds of GCC, such as Cygwin
2471    version 1.1.8, using the default configure settings.  It appears to be
2472    necessary to specify the -mwin32 flag when compiling, and define
2473    __MSVCRT__, like so:
2474    
2475  * On a Sun running SunOS 4.1.1, you get this error message from GNU ld:    configure --with-gcc --cflags -mwin32 --cflags -D__MSVCRT__
2476    
2477      /lib/libc.a(_Q_sub.o): Undefined symbol __Q_get_rp_rd referenced from text segment  *** Building the MS-Windows port fails with a CreateProcess failure.
2478    
2479  The problem is in the Sun shared C library, not in GNU ld.  Some versions of mingw32 make on some versions of Windows do not seem
2480    to detect the shell correctly. Try "make SHELL=cmd.exe", or if that
2481    fails, try running make from Cygwin bash instead.
2482    
2483  The solution is to install Patch-ID# 100267-03 from Sun.  *** Building the MS-Windows port with Leim fails in the `leim' directory.
2484    
2485  * Self documentation messages are garbled.  The error message might be something like this:
2486    
2487  This means that the file `etc/DOC-...' doesn't properly correspond   Converting d:/emacs-21.3/leim/CXTERM-DIC/4Corner.tit to quail-package...
2488  with the Emacs executable.  Redumping Emacs and then installing the   Invalid ENCODE: value in TIT dictionary
2489  corresponding pair of files should fix the problem.   NMAKE : fatal error U1077: '"../src/obj-spd/i386/emacs.exe"' : return code
2490           '0xffffffff'
2491     Stop.
2492    
2493  * Trouble using ptys on AIX.  This can happen if the Leim distribution is unpacked with a program
2494    which converts the `*.tit' files to DOS-style CR-LF text format.  The
2495    `*.tit' files in the leim/CXTERM-DIC directory require Unix-style line
2496    endings to compile properly, because Emacs reads them without any code
2497    or EOL conversions.
2498    
2499  People often install the pty devices on AIX incorrectly.  The solution is to make sure the program used to unpack Leim does not
2500  Use `smit pty' to reinstall them properly.  change the files' line endings behind your back.  The GNU FTP site has
2501    in the `/gnu/emacs/windows' directory a program called `djtarnt.exe'
2502    which can be used to unpack `.tar.gz' and `.zip' archives without
2503    mangling them.
2504    
2505  * Shell mode on HP/UX gives the message, "`tty`: Ambiguous".  *** Building `ctags' for MS-Windows with the MinGW port of GCC fails.
2506    
2507  christos@theory.tn.cornell.edu says:  This might happen due to a bug in the MinGW header assert.h, which
2508    defines the `assert' macro with a trailing semi-colon.  The following
2509    patch to assert.h should solve this:
2510    
2511  The problem is that in your .cshrc you have something that tries to  *** include/assert.h.orig       Sun Nov  7 02:41:36 1999
2512  execute `tty`. If you are not running the shell on a real tty then  --- include/assert.h    Mon Jan 29 11:49:10 2001
2513  tty will print "not a tty". Csh expects one word in some places,  ***************
2514  but tty is giving it back 3.  *** 41,47 ****
2515      /*
2516       * If not debugging, assert does nothing.
2517       */
2518    ! #define assert(x)     ((void)0);
2519    
2520  The solution is to add a pair of quotes around `tty` to make it a single    #else /* debugging enabled */
 word:  
2521    
2522  if (`tty` == "/dev/console")  --- 41,47 ----
2523      /*
2524       * If not debugging, assert does nothing.
2525       */
2526    ! #define assert(x)     ((void)0)
2527    
2528  should be changed to:    #else /* debugging enabled */
2529    
 if ("`tty`" == "/dev/console")  
2530    
2531  Even better, move things that set up terminal sections out of .cshrc  ** Linking
 and into .login.  
2532    
2533  * Using X Windows, control-shift-leftbutton makes Emacs hang.  *** Building Emacs with a system compiler fails to link because of an
2534    undefined symbol such as __eprintf which does not appear in Emacs.
2535    
2536  Use the shell command `xset bc' to make the old X Menu package work.  This can happen if some of the libraries linked into Emacs were built
2537    with GCC, but Emacs itself is being linked with a compiler other than
2538    GCC.  Object files compiled with GCC might need some helper functions
2539    from libgcc.a, the library which comes with GCC, but the system
2540    compiler does not instruct the linker to search libgcc.a during the
2541    link stage.
2542    
2543  * Emacs running under X Windows does not handle mouse clicks.  A solution is to link with GCC, like this:
 * `emacs -geometry 80x20' finds a file named `80x20'.  
2544    
2545  One cause of such problems is having (setq term-file-prefix nil) in          make CC=gcc
 your .emacs file.  Another cause is a bad value of EMACSLOADPATH in  
 the environment.  
2546    
2547  * Emacs gets error message from linker on Sun.  Since the .o object files already exist, this will not recompile Emacs
2548    with GCC, but just restart by trying again to link temacs.
2549    
2550  If the error message says that a symbol such as `f68881_used' or  *** AIX 1.3 ptf 0013: Link failure.
 `ffpa_used' or `start_float' is undefined, this probably indicates  
 that you have compiled some libraries, such as the X libraries,  
 with a floating point option other than the default.  
2551    
2552  It's not terribly hard to make this work with small changes in  There is a real duplicate definition of the function `_slibc_free' in
2553  crt0.c together with linking with Fcrt1.o, Wcrt1.o or Mcrt1.o.  the library /lib/libc_s.a (just do nm on it to verify).  The
2554  However, the easiest approach is to build Xlib with the default  workaround/fix is:
 floating point option: -fsoft.  
2555    
2556  * Emacs fails to get default settings from X Windows server.      cd /lib
2557        ar xv libc_s.a NLtmtime.o
2558        ar dv libc_s.a NLtmtime.o
2559    
2560  The X library in X11R4 has a bug; it interchanges the 2nd and 3rd  *** AIX 4.1.2: Linker error messages such as
2561  arguments to XGetDefaults.  Define the macro XBACKWARDS in config.h to    ld: 0711-212 SEVERE ERROR: Symbol .__quous, found in the global symbol table
2562  tell Emacs to compensate for this.          of archive /usr/lib/libIM.a, was not defined in archive member shr.o.
2563    
2564  I don't believe there is any way Emacs can determine for itself  This is a problem in libIM.a.  You can work around it by executing
2565  whether this problem is present on a given system.  these shell commands in the src subdirectory of the directory where
2566    you build Emacs:
2567    
2568  * Keyboard input gets confused after a beep when using a DECserver      cp /usr/lib/libIM.a .
2569    as a concentrator.      chmod 664 libIM.a
2570        ranlib libIM.a
2571    
2572  This problem seems to be a matter of configuring the DECserver to use  Then change -lIM to ./libIM.a in the command to link temacs (in
2573  7 bit characters rather than 8 bit characters.  Makefile).
2574    
2575  * M-x shell persistently reports "Process shell exited abnormally with code 1".  *** Sun with acc: Link failure when using acc on a Sun.
2576    
2577  This happened on Suns as a result of what is said to be a bug in Sunos  To use acc, you need additional options just before the libraries, such as
 version 4.0.x.  The only fix was to reboot the machine.  
2578    
2579  * Programs running under terminal emulator do not recognize `emacs'     /usr/lang/SC2.0.1/values-Xt.o -L/usr/lang/SC2.0.1/cg87 -L/usr/lang/SC2.0.1
   terminal type.  
2580    
2581  The cause of this is a shell startup file that sets the TERMCAP  and you need to add -lansi just before -lc.
 environment variable.  The terminal emulator uses that variable to  
 provide the information on the special terminal type that Emacs  
 emulates.  
2582    
2583  Rewrite your shell startup file so that it does not change TERMCAP  The precise file names depend on the compiler version, so we
2584  in such a case.  You could use the following conditional which sets  cannot easily arrange to supply them.
 it only if it is undefined.  
2585    
2586      if ( ! ${?TERMCAP} ) setenv TERMCAP ~/my-termcap-file  *** Linking says that the functions insque and remque are undefined.
2587    
2588  Or you could set TERMCAP only when you set TERM--which should not  Change oldXMenu/Makefile by adding insque.o to the variable OBJS.
 happen in a non-login shell.  
2589    
2590  * X Windows doesn't work if DISPLAY uses a hostname.  *** `tparam' reported as a multiply-defined symbol when linking with ncurses.
2591    
2592  People have reported kernel bugs in certain systems that cause Emacs  This problem results from an incompatible change in ncurses, in
2593  not to work with X Windows if DISPLAY is set using a host name.  But  version 1.9.9e approximately.  This version is unable to provide a
2594  the problem does not occur if DISPLAY is set to `unix:0.0'.  I think  definition of tparm without also defining tparam.  This is also
2595  the bug has to do with SIGIO or FIONREAD.  incompatible with Terminfo; as a result, the Emacs Terminfo support
2596    does not work with this version of ncurses.
2597    
2598  You may be able to compensate for the bug by doing (set-input-mode nil nil).  The fix is to install a newer version of ncurses, such as version 4.2.
 However, that has the disadvantage of turning off interrupts, so that  
 you are unable to quit out of a Lisp program by typing C-g.  
2599    
2600  The easy way to do this is to put  ** Dumping
2601    
2602    (setq x-sigio-bug t)  *** Linux: Segfault during `make bootstrap' under certain recent versions of the Linux kernel.
2603    
2604  in your site-init.el file.  With certain recent Linux kernels (like the one of Redhat Fedora Core
2605    1), the new "Exec-shield" functionality is enabled by default, which
2606    creates a different memory layout that breaks the emacs dumper.
2607    
2608  * Problem with remote X server on Suns.  You can check the Exec-shield state like this:
2609    
2610  On a Sun, running Emacs on one machine with the X server on another      cat /proc/sys/kernel/exec-shield
 may not work if you have used the unshared system libraries.  This  
 is because the unshared libraries fail to use YP for host name lookup.  
 As a result, the host name you specify may not be recognized.  
2611    
2612  * Shell mode ignores interrupts on Apollo Domain  It returns 1 or 2 when Exec-shield is enabled, 0 otherwise.  Please
2613    read your system documentation for more details on Exec-shield and
2614    associated commands.
2615    
2616  You may find that M-x shell prints the following message:  When Exec-shield is enabled, building Emacs will segfault during the
2617    execution of this command:
2618    
2619     Warning: no access to tty; thus no job control in this shell...  temacs --batch --load loadup [dump|bootstrap]
2620    
2621  This can happen if there are not enough ptys on your system.  To work around this problem, it is necessary to temporarily disable
2622  Here is how to make more of them.  Exec-shield while building Emacs, using the `setarch' command like
2623    this:
2624    
2625      % cd /dev      setarch i386 ./configure <configure parameters>
2626      % ls pty*      setarch i386 make <make parameters>
     # shows how many pty's you have. I had 8, named pty0 to pty7)  
     % /etc/crpty 8  
     # creates eight new pty's  
2627    
2628  * Fatal signal in the command  temacs -l loadup inc dump  *** Fatal signal in the command  temacs -l loadup inc dump.
2629    
2630  This command is the final stage of building Emacs.  It is run by the  This command is the final stage of building Emacs.  It is run by the
2631  Makefile in the src subdirectory, or by build.com on VMS.  Makefile in the src subdirectory, or by build.com on VMS.
# Line 2927  Makefile in the src subdirectory, or by Line 2633  Makefile in the src subdirectory, or by
2633  It has been known to get fatal errors due to insufficient swapping  It has been known to get fatal errors due to insufficient swapping
2634  space available on the machine.  space available on the machine.
2635    
2636  On 68000's, it has also happened because of bugs in the  On 68000s, it has also happened because of bugs in the
2637  subroutine `alloca'.  Verify that `alloca' works right, even  subroutine `alloca'.  Verify that `alloca' works right, even
2638  for large blocks (many pages).  for large blocks (many pages).
2639    
2640  * test-distrib says that the distribution has been clobbered  *** test-distrib says that the distribution has been clobbered.
2641  * or, temacs prints "Command key out of range 0-127"  *** or, temacs prints "Command key out of range 0-127".
2642  * or, temacs runs and dumps emacs, but emacs totally fails to work.  *** or, temacs runs and dumps emacs, but emacs totally fails to work.
2643  * or, temacs gets errors dumping emacs  *** or, temacs gets errors dumping emacs.
2644    
2645  This can be because the .elc files have been garbled.  Do not be  This can be because the .elc files have been garbled.  Do not be
2646  fooled by the fact that most of a .elc file is text: these are  fooled by the fact that most of a .elc file is text: these are
# Line 2967  nonprinting characters, you can fix them Line 2673  nonprinting characters, you can fix them
2673    and remake temacs.    and remake temacs.
2674   7) Remake emacs.  It should work now, with valid .elc files.   7) Remake emacs.  It should work now, with valid .elc files.
2675    
2676  * temacs prints "Pure Lisp storage exhausted"  *** temacs prints "Pure Lisp storage exhausted".
2677    
2678  This means that the Lisp code loaded from the .elc and .el  This means that the Lisp code loaded from the .elc and .el
2679  files during  temacs -l loadup inc dump  took up more  files during  temacs -l loadup inc dump  took up more
# Line 2996  But in some of the cases listed above, t Line 2702  But in some of the cases listed above, t
2702  of something else that is wrong.  Be sure to check and fix the real  of something else that is wrong.  Be sure to check and fix the real
2703  problem.  problem.
2704    
2705  * Changes made to .el files do not take effect.  *** Linux: Emacs crashes when dumping itself on Mac PPC running Yellow Dog GNU/Linux.
2706    
2707  You may have forgotten to recompile them into .elc files.  The crashes happen inside the function Fmake_symbol; here's a typical
2708  Then the old .elc files will be loaded, and your changes  C backtrace printed by GDB:
 will not be seen.  To fix this, do M-x byte-recompile-directory  
 and specify the directory that contains the Lisp files.  
2709    
2710  Emacs should print a warning when loading a .elc file which is older    0x190c0c0 in Fmake_symbol ()
2711  than the corresponding .el file.    (gdb) where
2712      #0  0x190c0c0 in Fmake_symbol ()
2713      #1  0x1942ca4 in init_obarray ()
2714      #2  0x18b3500 in main ()
2715      #3  0x114371c in __libc_start_main (argc=5, argv=0x7ffff5b4, envp=0x7ffff5cc,
2716    
2717    This could happen because GCC version 2.95 and later changed the base
2718    of the load address to 0x10000000.  Emacs needs to be told about this,
2719    but we currently cannot do that automatically, because that breaks
2720    other versions of GNU/Linux on the MacPPC.  Until we find a way to
2721    distinguish between the Yellow Dog and the other varieties of
2722    GNU/Linux systems on the PPC, you will have to manually uncomment the
2723    following section near the end of the file src/m/macppc.h in the Emacs
2724    distribution:
2725    
2726      #if 0  /* This breaks things on PPC GNU/Linux except for Yellowdog,
2727                even with identical GCC, as, ld.  Let's take it out until we
2728                know what's really going on here.  */
2729      /* GCC 2.95 and newer on GNU/Linux PPC changed the load address to
2730         0x10000000.  */
2731      #if defined __linux__
2732      #if __GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 95)
2733      #define DATA_SEG_BITS  0x10000000
2734      #endif
2735      #endif
2736      #endif /* 0 */
2737    
2738  * The dumped Emacs crashes when run, trying to write pure data.  Remove the "#if 0" and "#endif" directives which surround this, save
2739    the file, and then reconfigure and rebuild Emacs.  The dumping process
2740    should now succeed.
2741    
2742    *** HPUX 10.20: Emacs crashes during dumping on the HPPA machine.
2743    
2744    This seems to be due to a GCC bug; it is fixed in GCC 2.8.1.
2745    
2746    ** Installation
2747    
2748    *** Installing Emacs gets an error running `install-info'.
2749    
2750    You need to install a recent version of Texinfo; that package
2751    supplies the `install-info' command.
2752    
2753    ** First execution
2754    
2755    *** Emacs binary is not in executable format, and cannot be run.
2756    
2757    This was reported to happen when Emacs is built in a directory mounted
2758    via NFS, for some combinations of NFS client and NFS server.
2759    Usually, the file `emacs' produced in these cases is full of
2760    binary null characters, and the `file' utility says:
2761    
2762        emacs: ASCII text, with no line terminators
2763    
2764    We don't know what exactly causes this failure.  A work-around is to
2765    build Emacs in a directory on a local disk.
2766    
2767    *** The dumped Emacs crashes when run, trying to write pure data.
2768    
2769  Two causes have been seen for such problems.  Two causes have been seen for such problems.
2770    
# Line 3021  of its files pure after dumping, but the Line 2779  of its files pure after dumping, but the
2779  not initialized are not supposed to be pure.  On these systems you  not initialized are not supposed to be pure.  On these systems you
2780  may need to add "#define static" to the m- or the s- file.  may need to add "#define static" to the m- or the s- file.
2781    
2782  * Compilation errors on VMS.  * Emacs 19 problems
2783    
2784  You will get warnings when compiling on VMS because there are  ** Error messages `Wrong number of arguments: #<subr where-is-internal>, 5'.
 variable names longer than 32 (or whatever it is) characters.  
 This is not an error.  Ignore it.  
2785    
2786  VAX C does not support #if defined(foo).  Uses of this construct  This typically results from having the powerkey library loaded.
2787  were removed, but some may have crept back in.  They must be rewritten.  Powerkey was designed for Emacs 19.22.  It is obsolete now because
2788    Emacs 19 now has this feature built in; and powerkey also calls
2789    where-is-internal in an obsolete way.
2790    
2791  There is a bug in the C compiler which fails to sign extend characters  So the fix is to arrange not to load powerkey.
 in conditional expressions.  The bug is:  
         char c = -1, d = 1;  
         int i;  
2792    
2793          i = d ? c : d;  * Runtime problems on legacy systems
 The result is i == 255;  the fix is to typecast the char in the  
 conditional expression as an (int).  Known occurrences of such  
 constructs in Emacs have been fixed.  
2794    
2795  * rmail gets error getting new mail  This section covers bugs reported on very old hardware or software.
2796    If you are using hardware and an operating system shipped after 2000,
2797    it is unlikely you will see any of these.
2798    
2799  rmail gets new mail from /usr/spool/mail/$USER using a program  ** Ancient operating systems
 called `movemail'.  This program interlocks with /bin/mail using  
 the protocol defined by /bin/mail.  
2800    
2801  There are two different protocols in general use.  One of them uses  *** ISC Unix
 the `flock' system call.  The other involves creating a lock file;  
 `movemail' must be able to write in /usr/spool/mail in order to do  
 this.  You control which one is used by defining, or not defining,  
 the macro MAIL_USE_FLOCK in config.h or the m- or s- file it includes.  
 IF YOU DON'T USE THE FORM OF INTERLOCKING THAT IS NORMAL ON YOUR  
 SYSTEM, YOU CAN LOSE MAIL!  
2802    
2803  If your system uses the lock file protocol, and fascist restrictions  **** ISC: display-time causes kernel problems on ISC systems.
 prevent ordinary users from writing the lock files in /usr/spool/mail,  
 you may need to make `movemail' setgid to a suitable group such as  
 `mail'.  You can use these commands (as root):  
2804    
2805          chgrp mail movemail  Under Interactive Unix versions 3.0.1 and 4.0 (and probably other
2806          chmod 2755 movemail  versions), display-time causes the loss of large numbers of STREVENT
2807    cells.  Eventually the kernel's supply of these cells is exhausted.
2808    This makes emacs and the whole system run slow, and can make other
2809    processes die, in particular pcnfsd.
2810    
2811  If your system uses the lock file protocol, and fascist restrictions  Other emacs functions that communicate with remote processes may have
2812  prevent ordinary users from writing the lock files in /usr/spool/mail,  the same problem.  Display-time seems to be far the worst.
 you may need to make `movemail' setgid to a suitable group such as  
 `mail'.  To do this,  use the following commands (as root) after doing the  
 make install.  
2813    
2814          chgrp mail movemail  The only known fix: Don't run display-time.
         chmod 2755 movemail  
2815    
2816  Installation normally copies movemail from the build directory to an  *** SunOS
 installation directory which is usually under /usr/local/lib.  The  
 installed copy of movemail is usually in the directory  
 /usr/local/lib/emacs/VERSION/TARGET.  You must change the group and  
 mode of the installed copy; changing the group and mode of the build  
 directory copy is ineffective.  
2817    
2818  * Emacs spontaneously displays "I-search: " at the bottom of the screen.  **** Sun 4.0.x: M-x shell persistently reports "Process shell exited abnormally with code 1".
2819    
2820  This means that Control-S/Control-Q (XON/XOFF) "flow control" is being  This happened on Suns as a result of what is said to be a bug in Sunos
2821  used.  C-s/C-q flow control is bad for Emacs editors because it takes  version 4.0.x.  The only fix was to reboot the machine.
 away C-s and C-q as user commands.  Since editors do not output long  
 streams of text without user commands, there is no need for a  
 user-issuable "stop output" command in an editor; therefore, a  
 properly designed flow control mechanism would transmit all possible  
 input characters without interference.  Designing such a mechanism is  
 easy, for a person with at least half a brain.  
2822    
2823  There are three possible reasons why flow control could be taking place:  **** SunOS4.1.1 and SunOS4.1.3: Mail is lost when sent to local aliases.
2824    
2825    1) Terminal has not been told to disable flow control  Many emacs mail user agents (VM and rmail, for instance) use the
2826    2) Insufficient padding for the terminal in use  sendmail.el library.  This library can arrange for mail to be
2827    3) Some sort of terminal concentrator or line switch is responsible  delivered by passing messages to the /usr/lib/sendmail (usually)
2828    program .  In doing so, it passes the '-t' flag to sendmail, which
2829    means that the name of the recipient of the message is not on the
2830    command line and, therefore, that sendmail must parse the message to
2831    obtain the destination address.
2832    
2833  First of all, many terminals have a set-up mode which controls whether  There is a bug in the SunOS4.1.1 and SunOS4.1.3 versions of sendmail.
2834  they generate XON/XOFF flow control characters.  This must be set to  In short, when given the -t flag, the SunOS sendmail won't recognize
2835  "no XON/XOFF" in order for Emacs to work.  Sometimes there is an  non-local (i.e. NIS) aliases.  It has been reported that the Solaris
2836  escape sequence that the computer can send to turn flow control off  2.x versions of sendmail do not have this bug.  For those using SunOS
2837  and on.  If so, perhaps the termcap `ti' string should turn flow  4.1, the best fix is to install sendmail V8 or IDA sendmail (which
2838  control off, and the `te' string should turn it on.  have other advantages over the regular sendmail as well).  At the time
2839    of this writing, these official versions are available:
2840    
2841  Once the terminal has been told "no flow control", you may find it   Sendmail V8 on ftp.cs.berkeley.edu in /ucb/sendmail:
2842  needs more padding.  The amount of padding Emacs sends is controlled     sendmail.8.6.9.base.tar.Z (the base system source & documentation)
2843  by the termcap entry for the terminal in use, and by the output baud     sendmail.8.6.9.cf.tar.Z   (configuration files)
2844  rate as known by the kernel.  The shell command `stty' will print     sendmail.8.6.9.misc.tar.Z (miscellaneous support programs)
2845  your output baud rate; `stty' with suitable arguments will set it if     sendmail.8.6.9.xdoc.tar.Z (extended documentation, with postscript)
 it is wrong.  Setting to a higher speed causes increased padding.  If  
 the results are wrong for the correct speed, there is probably a  
 problem in the termcap entry.  You must speak to a local Unix wizard  
 to fix this.  Perhaps you are just using the wrong terminal type.  
2846    
2847  For terminals that lack a "no flow control" mode, sometimes just   IDA sendmail on vixen.cso.uiuc.edu in /pub:
2848  giving lots of padding will prevent actual generation of flow control     sendmail-5.67b+IDA-1.5.tar.gz
 codes.  You might as well try it.  
2849    
2850  If you are really unlucky, your terminal is connected to the computer  **** Sunos 5.3: Subprocesses remain, hanging but not zombies.
 through a concentrator which sends XON/XOFF flow control to the  
 computer, or it insists on sending flow control itself no matter how  
 much padding you give it.  Unless you can figure out how to turn flow  
 control off on this concentrator (again, refer to your local wizard),  
 you are screwed!  You should have the terminal or concentrator  
 replaced with a properly designed one.  In the mean time, some drastic  
 measures can make Emacs semi-work.  
2851    
2852  You can make Emacs ignore C-s and C-q and let the operating system  A bug in Sunos 5.3 causes Emacs subprocesses to remain after Emacs
2853  handle them.  To do this on a per-session basis, just type M-x  exits.  Sun patch # 101415-02 is part of the fix for this, but it only
2854  enable-flow-control RET.  You will see a message that C-\ and C-^ are  applies to ptys, and doesn't fix the problem with subprocesses
2855  now translated to C-s and C-q.  (Use the same command M-x  communicating through pipes.
 enable-flow-control to turn *off* this special mode.  It toggles flow  
 control handling.)  
2856    
2857  If C-\ and C-^ are inconvenient for you (for example, if one of them  **** Sunos 4: You get the error ld: Undefined symbol __lib_version.
 is the escape character of your terminal concentrator), you can choose  
 other characters by setting the variables flow-control-c-s-replacement  
 and flow-control-c-q-replacement.  But choose carefully, since all  
 other control characters are already used by emacs.  
2858    
2859  IMPORTANT: if you type C-s by accident while flow control is enabled,  This is the result of using cc or gcc with the shared library meant
2860  Emacs output will freeze, and you will have to remember to type C-q in  for acc (the Sunpro compiler).  Check your LD_LIBRARY_PATH and delete
2861  order to continue.  /usr/lang/SC2.0.1 or some similar directory.
2862    
2863  If you work in an environment where a majority of terminals of a  **** SunOS 4.1.3: Emacs unpredictably crashes in _yp_dobind_soft.
 certain type are flow control hobbled, you can use the function  
 `enable-flow-control-on' to turn on this flow control avoidance scheme  
 automatically.  Here is an example:  
2864    
2865  (enable-flow-control-on "vt200" "vt300" "vt101" "vt131")  This happens if you configure Emacs specifying just `sparc-sun-sunos4'
2866    on a system that is version 4.1.3.  You must specify the precise
2867    version number (or let configure figure out the configuration, which
2868    it can do perfectly well for SunOS).
2869    
2870  If this isn't quite correct (e.g. you have a mixture of flow-control hobbled  **** Sunos 4.1.3: Emacs gets hung shortly after startup.
 and good vt200 terminals), you can still run enable-flow-control  
 manually.  
2871    
2872  I have no intention of ever redesigning the Emacs command set for the  We think this is due to a bug in Sunos.  The word is that
2873  assumption that terminals use C-s/C-q flow control.  XON/XOFF flow  one of these Sunos patches fixes the bug:
 control technique is a bad design, and terminals that need it are bad  
 merchandise and should not be purchased.  Now that X is becoming  
 widespread, XON/XOFF seems to be on the way out.  If you can get some  
 use out of GNU Emacs on inferior terminals, more power to you, but I  
 will not make Emacs worse for properly designed systems for the sake  
 of inferior systems.  
2874    
2875  * Control-S and Control-Q commands are ignored completely.  100075-11  100224-06  100347-03  100482-05  100557-02  100623-03  100804-03  101080-01
2876    100103-12  100249-09             100496-02  100564-07  100630-02  100891-10  101134-01
2877    100170-09  100296-04  100377-09  100507-04  100567-04  100650-02  101070-01  101145-01
2878    100173-10  100305-15  100383-06  100513-04  100570-05  100689-01  101071-03  101200-02
2879    100178-09  100338-05  100421-03  100536-02  100584-05  100784-01  101072-01  101207-01
2880    
2881  For some reason, your system is using brain-damaged C-s/C-q flow  We don't know which of these patches really matter.  If you find out
2882  control despite Emacs's attempts to turn it off.  Perhaps your  which ones, please inform bug-gnu-emacs@gnu.org.
 terminal is connected to the computer through a concentrator  
 that wants to use flow control.  
2883    
2884  You should first try to tell the concentrator not to use flow control.  **** SunOS 4: Emacs processes keep going after you kill the X server
2885  If you succeed in this, try making the terminal work without  (or log out, if you logged in using X).
 flow control, as described in the preceding section.  
2886    
2887  If that line of approach is not successful, map some other characters  Someone reported that recompiling with GCC 2.7.0 fixed this problem.
 into C-s and C-q using keyboard-translate-table.  The example above  
 shows how to do this with C-^ and C-\.  
2888    
2889  * Control-S and Control-Q commands are ignored completely on a net connection.  **** SunOS: You get linker errors
2890       ld: Undefined symbol
2891          _get_wmShellWidgetClass
2892          _get_applicationShellWidgetClass
2893    
2894  Some versions of rlogin (and possibly telnet) do not pass flow  The fix to this is to install patch 100573 for OpenWindows 3.0
2895  control characters to the remote system to which they connect.  or link libXmu statically.
 On such systems, emacs on the remote system cannot disable flow  
 control on the local system.  
2896    
2897  One way to cure this is to disable flow control on the local host  *** Apollo Domain
 (the one running rlogin, not the one running rlogind) using the  
 stty command, before starting the rlogin process.  On many systems,  
 "stty start u stop u" will do this.  
2898    
2899  Some versions of tcsh will prevent even this from working.  One way  **** Shell mode ignores interrupts on Apollo Domain.
 around this is to start another shell before starting rlogin, and  
 issue the stty command to disable flow control from that shell.  
2900    
2901  If none of these methods work, the best solution is to type  You may find that M-x shell prints the following message:
 M-x enable-flow-control at the beginning of your emacs session, or  
 if you expect the problem to continue, add a line such as the  
 following to your .emacs (on the host running rlogind):  
2902    
2903  (enable-flow-control-on "vt200" "vt300" "vt101" "vt131")     Warning: no access to tty; thus no job control in this shell...
2904    
2905  See the entry about spontaneous display of I-search (above) for more  This can happen if there are not enough ptys on your system.
2906  info.  Here is how to make more of them.
2907    
2908  * Screen is updated wrong, but only on one kind of terminal.      % cd /dev
2909        % ls pty*
2910        # shows how many pty's you have. I had 8, named pty0 to pty7)
2911        % /etc/crpty 8
2912        # creates eight new pty's
2913    
2914  This could mean that the termcap entry you are using for that  *** Irix
 terminal is wrong, or it could mean that Emacs has a bug handing  
 the combination of features specified for that terminal.  
2915    
2916  The first step in tracking this down is to record what characters  *** Irix 6.2: No visible display on mips-sgi-irix6.2 when compiling with GCC 2.8.1.
 Emacs is sending to the terminal.  Execute the Lisp expression  
 (open-termscript "./emacs-script") to make Emacs write all  
 terminal output into the file ~/emacs-script as well; then do  
 what makes the screen update wrong, and look at the file  
 and decode the characters using the manual for the terminal.  
 There are several possibilities:  
2917    
2918  1) The characters sent are correct, according to the terminal manual.  This problem went away after installing the latest IRIX patches
2919    as of 8 Dec 1998.
2920    
2921  In this case, there is no obvious bug in Emacs, and most likely you  The same problem has been reported on Irix 6.3.
 need more padding, or possibly the terminal manual is wrong.  
2922    
2923  2) The characters sent are incorrect, due to an obscure aspect  *** Irix 6.3: substituting environment variables in file names
2924   of the terminal behavior not described in an obvious way  in the minibuffer gives peculiar error messages such as
  by termcap.  
2925    
2926  This case is hard.  It will be necessary to think of a way for     Substituting nonexistent environment variable ""
 Emacs to distinguish between terminals with this kind of behavior  
 and other terminals that behave subtly differently but are  
 classified the same by termcap; or else find an algorithm for  
 Emacs to use that avoids the difference.  Such changes must be  
 tested on many kinds of terminals.  
2927    
2928  3) The termcap entry is wrong.  This is not an Emacs bug; it is caused by something in SGI patch
2929    003082 August 11, 1998.
2930    
2931  See the file etc/TERMS for information on changes  *** OPENSTEP
 that are known to be needed in commonly used termcap entries  
 for certain terminals.  
2932    
2933  4) The characters sent are incorrect, and clearly cannot be  **** OPENSTEP 4.2: Compiling syntax.c with gcc 2.7.2.1 fails.
  right for any terminal with the termcap entry you were using.  
2934    
2935  This is unambiguously an Emacs bug, and can probably be fixed  The compiler was reported to crash while compiling syntax.c with the
2936  in termcap.c, tparam.c, term.c, scroll.c, cm.c or dispnew.c.  following message:
2937    
2938  * Output from Control-V is slow.     cc: Internal compiler error: program cc1obj got fatal signal 11
2939    
2940  On many bit-map terminals, scrolling operations are fairly slow.  To work around this, replace the macros UPDATE_SYNTAX_TABLE_FORWARD,
2941  Often the termcap entry for the type of terminal in use fails  INC_BOTH, and INC_FROM with functions.  To this end, first define 3
2942  to inform Emacs of this.  The two lines at the bottom of the screen  functions, one each for every macro.  Here's an example:
 before a Control-V command are supposed to appear at the top after  
 the Control-V command.  If Emacs thinks scrolling the lines is fast,  
 it will scroll them to the top of the screen.  
2943    
2944  If scrolling is slow but Emacs thinks it is fast, the usual reason is      static int update_syntax_table_forward(int from)
2945  that the termcap entry for the terminal you are using does not      {
2946  specify any padding time for the `al' and `dl' strings.  Emacs          return(UPDATE_SYNTAX_TABLE_FORWARD(from));
2947  concludes that these operations take only as much time as it takes to      }/*update_syntax_table_forward*/
 send the commands at whatever line speed you are using.  You must  
 fix the termcap entry to specify, for the `al' and `dl', as much  
 time as the operations really take.  
2948    
2949  Currently Emacs thinks in terms of serial lines which send characters  Then replace all references to UPDATE_SYNTAX_TABLE_FORWARD in syntax.c
2950  at a fixed rate, so that any operation which takes time for the  with a call to the function update_syntax_table_forward.
 terminal to execute must also be padded.  With bit-map terminals  
 operated across networks, often the network provides some sort of  
 flow control so that padding is never needed no matter how slow  
 an operation is.  You must still specify a padding time if you want  
 Emacs to realize that the operation takes a long time.  This will  
 cause padding characters to be sent unnecessarily, but they do  
 not really cost much.  They will be transmitted while the scrolling  
 is happening and then discarded quickly by the terminal.  
2951    
2952  Most bit-map terminals provide commands for inserting or deleting  *** Solaris 2.x
 multiple lines at once.  Define the `AL' and `DL' strings in the  
 termcap entry to say how to do these things, and you will have  
 fast output without wasted padding characters.  These strings should  
 each contain a single %-spec saying how to send the number of lines  
 to be scrolled.  These %-specs are like those in the termcap  
 `cm' string.  
2953    
2954  You should also define the `IC' and `DC' strings if your terminal  **** Strange results from format %d in a few cases, on a Sun.
 has a command to insert or delete multiple characters.  These  
 take the number of positions to insert or delete as an argument.  
2955    
2956  A `cs' string to set the scrolling region will reduce the amount  Sun compiler version SC3.0 has been found to miscompile part of
2957  of motion you see on the screen when part of the screen is scrolled.  editfns.c.  The workaround is to compile with some other compiler such
2958    as GCC.
2959    
2960  * Your Delete key sends a Backspace to the terminal, using an AIXterm.  **** On Solaris, Emacs dumps core if lisp-complete-symbol is called.
2961    
2962  The solution is to include in your .Xdefaults the lines:  If you compile Emacs with the -fast or -xO4 option with version 3.0.2
2963    of the Sun C compiler, Emacs dumps core when lisp-complete-symbol is
2964    called.  The problem does not happen if you compile with GCC.
2965    
2966     *aixterm.Translations: #override <Key>BackSpace: string(0x7f)  **** On Solaris, Emacs crashes if you use (display-time).
    aixterm*ttyModes: erase ^?  
2967    
2968  This makes your Backspace key send DEL (ASCII 127).  This can happen if you configure Emacs without specifying the precise
2969    version of Solaris that you are using.
2970    
2971  * You type Control-H (Backspace) expecting to delete characters.  **** Solaris 2.3 and 2.4: Unpredictable segmentation faults.
2972    
2973  Put `stty dec' in your .login file and your problems will disappear  A user reported that this happened in 19.29 when it was compiled with
2974  after a day or two.  the Sun compiler, but not when he recompiled with GCC 2.7.0.
2975    
2976  The choice of Backspace for erasure was based on confusion, caused by  We do not know whether something in Emacs is partly to blame for this.
 the fact that backspacing causes erasure (later, when you type another  
 character) on most display terminals.  But it is a mistake.  Deletion  
 of text is not the same thing as backspacing followed by failure to  
 overprint.  I do not wish to propagate this confusion by conforming  
 to it.  
2977    
2978  For this reason, I believe `stty dec' is the right mode to use,  **** Solaris 2.4: Emacs dumps core on startup.
 and I have designed Emacs to go with that.  If there were a thousand  
 other control characters, I would define Control-h to delete as well;  
 but there are not very many other control characters, and I think  
 that providing the most mnemonic possible Help character is more  
 important than adapting to people who don't use `stty dec'.  
2979    
2980  If you are obstinate about confusing buggy overprinting with deletion,  Bill Sebok says that the cause of this is Solaris 2.4 vendor patch
2981  you can redefine Backspace in your .emacs file:  102303-05, which extends the Solaris linker to deal with the Solaris
2982    (global-set-key "\b" 'delete-backward-char)  Common Desktop Environment's linking needs.  You can fix the problem
2983  You can probably access  help-command  via f1.  by removing this patch and installing patch 102049-02 instead.
2984    However, that linker version won't work with CDE.
2985    
2986  * Editing files through RFS gives spurious "file has changed" warnings.  Solaris 2.5 comes with a linker that has this bug.  It is reported that if
2987  It is possible that a change in Emacs 18.37 gets around this problem,  you install all the latest patches (as of June 1996), the bug is fixed.
2988  but in case not, here is a description of how to fix the RFS bug that  We suspect the crucial patch is one of these, but we don't know
2989  causes it.  for certain.
2990    
2991      There was a serious pair of bugs in the handling of the fsync() system          103093-03: [README] SunOS 5.5: kernel patch (2140557 bytes)
2992      call in the RFS server.          102832-01: [README] OpenWindows 3.5: Xview Jumbo Patch (4181613 bytes)
2993            103242-04: [README] SunOS 5.5: linker patch (595363 bytes)
2994    
2995      The first is that the fsync() call is handled as another name for the  (One user reports that the bug was fixed by those patches together
2996      close() system call (!!).  It appears that fsync() is not used by very  with patches 102980-04, 103279-01, 103300-02, and 103468-01.)
     many programs; Emacs version 18 does an fsync() before closing files  
     to make sure that the bits are on the disk.  
2997    
2998      This is fixed by the enclosed patch to the RFS server.  If you can determine which patch does fix the bug, please tell
2999    bug-gnu-emacs@gnu.org.
3000    
3001      The second, more serious problem, is that fsync() is treated as a  Meanwhile, the GNU linker links Emacs properly on both Solaris 2.4 and
3002      non-blocking system call (i.e., it's implemented as a message that  Solaris 2.5.
     gets sent to the remote system without waiting for a reply).  Fsync is  
     a useful tool for building atomic file transactions.  Implementing it  
     as a non-blocking RPC call (when the local call blocks until the sync  
     is done) is a bad idea; unfortunately, changing it will break the RFS  
     protocol.  No fix was supplied for this problem.  
3003    
3004      (as always, your line numbers may vary)  **** Solaris 2.4: Dired hangs and C-g does not work.  Or Emacs hangs
3005    forever waiting for termination of a subprocess that is a zombie.
3006    
3007      % rcsdiff -c -r1.2 serversyscall.c  casper@fwi.uva.nl says the problem is in X11R6.  Rebuild libX11.so
3008      RCS file: RCS/serversyscall.c,v  after changing the file xc/config/cf/sunLib.tmpl.  Change the lines
3009      retrieving revision 1.2  
3010      diff -c -r1.2 serversyscall.c      #if ThreadedX
3011      *** /tmp/,RCSt1003677   Wed Jan 28 15:15:02 1987      #define SharedX11Reqs -lthread
3012      --- serversyscall.c     Wed Jan 28 15:14:48 1987      #endif
3013      ***************  
3014      *** 163,169 ****  to:
3015              /*  
3016               * No return sent for close or fsync!      #if OSMinorVersion < 4
3017               */      #if ThreadedX
3018      !       if (syscall == RSYS_close || syscall == RSYS_fsync)      #define SharedX11Reqs -lthread
3019                      proc->p_returnval = deallocate_fd(proc, msg->m_args[0]);      #endif
3020              else      #endif
3021              {  
3022      --- 166,172 ----  Be sure also to edit x/config/cf/sun.cf so that OSMinorVersion is 4
3023              /*  (as it should be for Solaris 2.4).  The file has three definitions for
3024               * No return sent for close or fsync!  OSMinorVersion: the first is for x86, the second for SPARC under
3025               */  Solaris, and the third for SunOS 4.  Make sure to update the
3026      !       if (syscall == RSYS_close)  definition for your type of machine and system.
3027                      proc->p_returnval = deallocate_fd(proc, msg->m_args[0]);  
3028              else  Then do `make Everything' in the top directory of X11R6, to rebuild
3029              {  the makefiles and rebuild X.  The X built this way work only on
3030    Solaris 2.4, not on 2.3.
3031    
3032    For multithreaded X to work it is necessary to install patch
3033    101925-02 to fix problems in header files [2.4].  You need
3034    to reinstall gcc or re-run just-fixinc after installing that
3035    patch.
3036    
3037    However, Frank Rust <frust@iti.cs.tu-bs.de> used a simpler solution:
3038    he changed
3039        #define ThreadedX          YES
3040    to
3041        #define ThreadedX          NO
3042    in sun.cf and did `make World' to rebuild X11R6.  Removing all
3043    `-DXTHREAD*' flags and `-lthread' entries from lib/X11/Makefile and
3044    typing 'make install' in that directory also seemed to work.
3045    
3046    **** Solaris 2.x: GCC complains "64 bit integer types not supported".
3047    
3048    This suggests that GCC is not installed correctly.  Most likely you
3049    are using GCC 2.7.2.3 (or earlier) on Solaris 2.6 (or later); this
3050    does not work without patching.  To run GCC 2.7.2.3 on Solaris 2.6 or
3051    later, you must patch fixinc.svr4 and reinstall GCC from scratch as
3052    described in the Solaris FAQ
3053    <http://www.wins.uva.nl/pub/solaris/solaris2.html>.  A better fix is
3054    to upgrade to GCC 2.8.1 or later.
3055    
3056    **** Solaris 2.7: Building Emacs with WorkShop Compilers 5.0 98/12/15
3057    C 5.0 failed, apparently with non-default CFLAGS, most probably due to
3058    compiler bugs.  Using Sun Solaris 2.7 Sun WorkShop 6 update 1 C
3059    release was reported to work without problems.  It worked OK on
3060    another system with Solaris 8 using apparently the same 5.0 compiler
3061    and the default CFLAGS.
3062    
3063    **** Solaris 2.x: Emacs dumps core when built with Motif.
3064    
3065    The Solaris Motif libraries are buggy, at least up through Solaris 2.5.1.
3066    Install the current Motif runtime library patch appropriate for your host.
3067    (Make sure the patch is current; some older patch versions still have the bug.)
3068    You should install the other patches recommended by Sun for your host, too.
3069    You can obtain Sun patches from ftp://sunsolve.sun.com/pub/patches/;
3070    look for files with names ending in `.PatchReport' to see which patches
3071    are currently recommended for your host.
3072    
3073    On Solaris 2.6, Emacs is said to work with Motif when Solaris patch
3074    105284-12 is installed, but fail when 105284-15 is installed.
3075    105284-18 might fix it again.
3076    
3077    *** Solaris 2.6 and 7: the Compose key does not work.
3078    
3079    This is a bug in Motif in Solaris.  Supposedly it has been fixed for
3080    the next major release of Solaris.  However, if someone with Sun
3081    support complains to Sun about the bug, they may release a patch.
3082    If you do this, mention Sun bug #4188711.
3083    
3084    One workaround is to use a locale that allows non-ASCII characters.
3085    For example, before invoking emacs, set the LC_ALL environment
3086    variable to "en_US" (American English).  The directory /usr/lib/locale
3087    lists the supported locales; any locale other than "C" or "POSIX"
3088    should do.
3089    
3090    pen@lysator.liu.se says (Feb 1998) that the Compose key does work
3091    if you link with the MIT X11 libraries instead of the Solaris X11
3092    libraries.
3093    
3094    *** Ultrix and Digital Unix
3095    
3096    **** Ultrix 4.2: `make install' fails on install-doc with `Error 141'.
3097    
3098    This happens on Ultrix 4.2 due to failure of a pipeline of tar
3099    commands.  We don't know why they fail, but the bug seems not to be in
3100    Emacs.  The workaround is to run the shell command in install-doc by
3101    hand.
3102    
3103    **** Digital Unix 4.0: Garbled display on non-X terminals when Emacs runs.
3104    
3105    So far it appears that running `tset' triggers this problem (when TERM
3106    is vt100, at least).  If you do not run `tset', then Emacs displays
3107    properly.  If someone can tell us precisely which effect of running
3108    `tset' actually causes the problem, we may be able to implement a fix
3109    in Emacs.
3110    
3111    **** Ultrix: `expand-file-name' fails to work on any but the machine you dumped Emacs on.
3112    
3113    On Ultrix, if you use any of the functions which look up information
3114    in the passwd database before dumping Emacs (say, by using
3115    expand-file-name in site-init.el), then those functions will not work
3116    in the dumped Emacs on any host but the one Emacs was dumped on.
3117    
3118    The solution?  Don't use expand-file-name in site-init.el, or in
3119    anything it loads.  Yuck - some solution.
3120    
3121    I'm not sure why this happens; if you can find out exactly what is
3122    going on, and perhaps find a fix or a workaround, please let us know.
3123    Perhaps the YP functions cache some information, the cache is included
3124    in the dumped Emacs, and is then inaccurate on any other host.
3125    
3126    *** SVr4
3127    
3128    **** SVr4: On some variants of SVR4, Emacs does not work at all with X.
3129    
3130    Try defining BROKEN_FIONREAD in your config.h file.  If this solves
3131    the problem, please send a bug report to tell us this is needed; be
3132    sure to say exactly what type of machine and system you are using.
3133    
3134    **** SVr4: After running emacs once, subsequent invocations crash.
3135    
3136    Some versions of SVR4 have a serious bug in the implementation of the
3137    mmap () system call in the kernel; this causes emacs to run correctly
3138    the first time, and then crash when run a second time.
3139    
3140    Contact your vendor and ask for the mmap bug fix; in the mean time,
3141    you may be able to work around the problem by adding a line to your
3142    operating system description file (whose name is reported by the
3143    configure script) that reads:
3144    #define SYSTEM_MALLOC
3145    This makes Emacs use memory less efficiently, but seems to work around
3146    the kernel bug.
3147    
3148    *** Linux 1.x
3149    
3150    **** Linux 1.0-1.04: Typing C-c C-c in Shell mode kills your X server.
3151    
3152    This happens with Linux kernel 1.0 thru 1.04, approximately.  The workaround is
3153    to define SIGNALS_VIA_CHARACTERS in config.h and recompile Emacs.
3154    Newer Linux kernel versions don't have this problem.
3155    
3156    **** Linux 1.3: Output from subprocess (such as man or diff) is randomly
3157    truncated on GNU/Linux systems.
3158    
3159    This is due to a kernel bug which seems to be fixed in Linux version
3160    1.3.75.
3161    
3162    ** MS-DOS
3163    
3164    *** When compiling with DJGPP on MS-Windows NT, "config msdos" fails.
3165    
3166    If the error message is "VDM has been already loaded", this is because
3167    Windows has a program called `redir.exe' that is incompatible with a
3168    program by the same name supplied with DJGPP, which is used by
3169    config.bat.  To resolve this, move the DJGPP's `bin' subdirectory to
3170    the front of your PATH environment variable.
3171    
3172    *** When compiling with DJGPP on MS-Windows 95, Make fails for some targets
3173    like make-docfile.
3174    
3175    This can happen if long file name support (the setting of environment
3176    variable LFN) when Emacs distribution was unpacked and during
3177    compilation are not the same.  See the MSDOG section of INSTALL for
3178    the explanation of how to avoid this problem.
3179    
3180    *** Emacs compiled with DJGPP complains at startup:
3181    
3182      "Wrong type of argument: internal-facep, msdos-menu-active-face"
3183    
3184    This can happen if you define an environment variable `TERM'.  Emacs
3185    on MSDOS uses an internal terminal emulator which is disabled if the
3186    value of `TERM' is anything but the string "internal".  Emacs then
3187    works as if its terminal were a dumb glass teletype that doesn't
3188    support faces.  To work around this, arrange for `TERM' to be
3189    undefined when Emacs runs.  The best way to do that is to add an
3190    [emacs] section to the DJGPP.ENV file which defines an empty value for
3191    `TERM'; this way, only Emacs gets the empty value, while the rest of
3192    your system works as before.
3193    
3194    *** MS-DOS: Emacs crashes at startup.
3195    
3196    Some users report that Emacs 19.29 requires dpmi memory management,
3197    and crashes on startup if the system does not have it.  We don't yet
3198    know why this happens--perhaps these machines don't have enough real
3199    memory, or perhaps something is wrong in Emacs or the compiler.
3200    However, arranging to use dpmi support is a workaround.
3201    
3202    You can find out if you have a dpmi host by running go32 without
3203    arguments; it will tell you if it uses dpmi memory.  For more
3204    information about dpmi memory, consult the djgpp FAQ.  (djgpp
3205    is the GNU C compiler as packaged for MSDOS.)
3206    
3207    Compiling Emacs under MSDOS is extremely sensitive for proper memory
3208    configuration.  If you experience problems during compilation, consider
3209    removing some or all memory resident programs (notably disk caches)
3210    and make sure that your memory managers are properly configured.  See
3211    the djgpp faq for configuration hints.
3212    
3213    *** Emacs compiled with DJGPP for MS-DOS/MS-Windows cannot access files
3214    in the directory with the special name `dev' under the root of any
3215    drive, e.g. `c:/dev'.
3216    
3217    This is an unfortunate side-effect of the support for Unix-style
3218    device names such as /dev/null in the DJGPP runtime library.  A
3219    work-around is to rename the problem directory to another name.
3220    
3221    *** MS-DOS+DJGPP: Problems on MS-DOG if DJGPP v2.0 is used to compile Emacs.
3222    
3223    There are two DJGPP library bugs which cause problems:
3224    
3225      * Running `shell-command' (or `compile', or `grep') you get
3226        `Searching for program: permission denied (EACCES), c:/command.com';
3227      * After you shell to DOS, Ctrl-Break kills Emacs.
3228    
3229    To work around these bugs, you can use two files in the msdos
3230    subdirectory: `is_exec.c' and `sigaction.c'.  Compile them and link
3231    them into the Emacs executable `temacs'; then they will replace the
3232    incorrect library functions.
3233    
3234    *** MS-DOS: Emacs compiled for MSDOS cannot find some Lisp files, or other
3235    run-time support files, when long filename support is enabled.
3236    
3237    Usually, this problem will manifest itself when Emacs exits
3238    immediately after flashing the startup screen, because it cannot find
3239    the Lisp files it needs to load at startup.  Redirect Emacs stdout
3240    and stderr to a file to see the error message printed by Emacs.
3241    
3242    Another manifestation of this problem is that Emacs is unable to load
3243    the support for editing program sources in languages such as C and
3244    Lisp.
3245    
3246    This can happen if the Emacs distribution was unzipped without LFN
3247    support, thus causing long filenames to be truncated to the first 6
3248    characters and a numeric tail that Windows 95 normally attaches to it.
3249    You should unzip the files again with a utility that supports long
3250    filenames (such as djtar from DJGPP or InfoZip's UnZip program
3251    compiled with DJGPP v2).  The MSDOG section of the file INSTALL
3252    explains this issue in more detail.
3253    
3254    Another possible reason for such failures is that Emacs compiled for
3255    MSDOS is used on Windows NT, where long file names are not supported
3256    by this version of Emacs, but the distribution was unpacked by an
3257    unzip program that preserved the long file names instead of truncating
3258    them to DOS 8+3 limits.  To be useful on NT, the MSDOS port of Emacs
3259    must be unzipped by a DOS utility, so that long file names are
3260    properly truncated.
3261    
3262    ** Archaic window managers and toolkits
3263    
3264    *** OpenLook: Under OpenLook, the Emacs window disappears when you type M-q.
3265    
3266    Some versions of the Open Look window manager interpret M-q as a quit
3267    command for whatever window you are typing at.  If you want to use
3268    Emacs with that window manager, you should try to configure the window
3269    manager to use some other command.   You can disable the
3270    shortcut keys entirely by adding this line to ~/.OWdefaults:
3271    
3272        OpenWindows.WindowMenuAccelerators: False
3273    
3274    **** twm: A position you specified in .Xdefaults is ignored, using twm.
3275    
3276    twm normally ignores "program-specified" positions.
3277    You can tell it to obey them with this command in your `.twmrc' file:
3278    
3279      UsePPosition  "on"            #allow clients to request a position
3280    
3281    ** Bugs related to old DEC hardware
3282    
3283    *** The Compose key on a DEC keyboard does not work as Meta key.
3284    
3285    This shell command should fix it:
3286    
3287      xmodmap -e 'keycode 0xb1 = Meta_L'
3288    
3289    *** Keyboard input gets confused after a beep when using a DECserver
3290    as a concentrator.
3291    
3292    This problem seems to be a matter of configuring the DECserver to use
3293    7 bit characters rather than 8 bit characters.
3294    
3295    * Build problems on legacy systems
3296    
3297    ** BSD/386 1.0: --with-x-toolkit option configures wrong.
3298    
3299    This problem is due to bugs in the shell in version 1.0 of BSD/386.
3300    The workaround is to edit the configure file to use some other shell,
3301    such as bash.
3302    
3303    ** Digital Unix 4.0: Emacs fails to build, giving error message
3304         Invalid dimension for the charset-ID 160
3305    
3306    This is due to a bug or an installation problem in GCC 2.8.0.
3307    Installing a more recent version of GCC fixes the problem.
3308    
3309    ** Digital Unix 4.0: Failure in unexec while dumping emacs.
3310    
3311    This problem manifests itself as an error message
3312    
3313        unexec: Bad address, writing data section to ...
3314    
3315    The user suspects that this happened because his X libraries
3316    were built for an older system version,
3317    
3318        ./configure --x-includes=/usr/include --x-libraries=/usr/shlib
3319    
3320    made the problem go away.
3321    
3322    ** Sunos 4.1.1: there are errors compiling sysdep.c.
3323    
3324    If you get errors such as
3325    
3326        "sysdep.c", line 2017: undefined structure or union
3327        "sysdep.c", line 2017: undefined structure or union
3328        "sysdep.c", line 2019: nodename undefined
3329    
3330    This can result from defining LD_LIBRARY_PATH.  It is very tricky
3331    to use that environment variable with Emacs.  The Emacs configure
3332    script links many test programs with the system libraries; you must
3333    make sure that the libraries available to configure are the same
3334    ones available when you build Emacs.
3335    
3336    ** SunOS 4.1.1: You get this error message from GNU ld:
3337    
3338        /lib/libc.a(_Q_sub.o): Undefined symbol __Q_get_rp_rd referenced from text segment
3339    
3340    The problem is in the Sun shared C library, not in GNU ld.
3341    
3342    The solution is to install Patch-ID# 100267-03 from Sun.
3343    
3344    ** Sunos 4.1: Undefined symbols when linking using --with-x-toolkit.
3345    
3346    If you get the undefined symbols _atowc _wcslen, _iswprint, _iswspace,
3347    _iswcntrl, _wcscpy, and _wcsncpy, then you need to add -lXwchar after
3348    -lXaw in the command that links temacs.
3349    
3350    This problem seems to arise only when the international language
3351    extensions to X11R5 are installed.
3352    
3353    ** SunOS: Emacs gets error message from linker on Sun.
3354    
3355    If the error message says that a symbol such as `f68881_used' or
3356    `ffpa_used' or `start_float' is undefined, this probably indicates
3357    that you have compiled some libraries, such as the X libraries,
3358    with a floating point option other than the default.
3359    
3360    It's not terribly hard to make this work with small changes in
3361    crt0.c together with linking with Fcrt1.o, Wcrt1.o or Mcrt1.o.
3362    However, the easiest approach is to build Xlib with the default
3363    floating point option: -fsoft.
3364    
3365    ** SunOS: Undefined symbols _dlopen, _dlsym and/or _dlclose.
3366    
3367    If you see undefined symbols _dlopen, _dlsym, or _dlclose when linking
3368    with -lX11, compile and link against the file mit/util/misc/dlsym.c in
3369    the MIT X11R5 distribution.  Alternatively, link temacs using shared
3370    libraries with s/sunos4shr.h.  (This doesn't work if you use the X
3371    toolkit.)
3372    
3373    If you get the additional error that the linker could not find
3374    lib_version.o, try extracting it from X11/usr/lib/X11/libvim.a in
3375    X11R4, then use it in the link.
3376    
3377    ** VMS: Compilation errors on VMS.
3378    
3379    You will get warnings when compiling on VMS because there are
3380    variable names longer than 32 (or whatever it is) characters.
3381    This is not an error.  Ignore it.
3382    
3383    VAX C does not support #if defined(foo).  Uses of this construct
3384    were removed, but some may have crept back in.  They must be rewritten.
3385    
3386    There is a bug in the C compiler which fails to sign extend characters
3387    in conditional expressions.  The bug is:
3388            char c = -1, d = 1;
3389            int i;
3390    
3391            i = d ? c : d;
3392    The result is i == 255;  the fix is to typecast the char in the
3393    conditional expression as an (int).  Known occurrences of such
3394    constructs in Emacs have been fixed.
3395    
3396  * Vax C compiler bugs affecting Emacs.  ** Vax C compiler bugs affecting Emacs.
3397    
3398  You may get one of these problems compiling Emacs:  You may get one of these problems compiling Emacs:
3399    
# Line 3403  causes the problem to go away. Line 3426  causes the problem to go away.
3426  The `contents' field of a Lisp vector is an array of Lisp_Objects,  The `contents' field of a Lisp vector is an array of Lisp_Objects,
3427  so you may see the problem happening with indexed references to that.  so you may see the problem happening with indexed references to that.
3428    
3429  * 68000 C compiler problems  ** 68000 C compiler problems
3430    
3431  Various 68000 compilers have different problems.  Various 68000 compilers have different problems.
3432  These are some that have been observed.  These are some that have been observed.
3433    
3434  ** Using value of assignment expression on union type loses.  *** Using value of assignment expression on union type loses.
3435  This means that  x = y = z;  or  foo (x = z);  does not work  This means that  x = y = z;  or  foo (x = z);  does not work
3436  if x is of type Lisp_Object.  if x is of type Lisp_Object.
3437    
3438  ** "cannot reclaim" error.  *** "cannot reclaim" error.
3439    
3440  This means that an expression is too complicated.  You get the correct  This means that an expression is too complicated.  You get the correct
3441  line number in the error message.  The code must be rewritten with  line number in the error message.  The code must be rewritten with
3442  simpler expressions.  simpler expressions.
3443    
3444  ** XCONS, XSTRING, etc macros produce incorrect code.  *** XCONS, XSTRING, etc macros produce incorrect code.
3445    
3446  If temacs fails to run at all, this may be the cause.  If temacs fails to run at all, this may be the cause.
3447  Compile this test program and look at the assembler code:  Compile this test program and look at the assembler code:
# Line 3438  In the XCONS, etc., macros in lisp.h you Line 3461  In the XCONS, etc., macros in lisp.h you
3461  This problem will not happen if the m-...h file for your type  This problem will not happen if the m-...h file for your type
3462  of machine defines NO_UNION_TYPE.  That is the recommended setting now.  of machine defines NO_UNION_TYPE.  That is the recommended setting now.
3463    
3464  * C compilers lose on returning unions  *** C compilers lose on returning unions.
3465    
3466  I hear that some C compilers cannot handle returning a union type.  I hear that some C compilers cannot handle returning a union type.
3467  Most of the functions in GNU Emacs return type Lisp_Object, which is  Most of the functions in GNU Emacs return type Lisp_Object, which is
# Line 3448  This problem will not happen if the m-.. Line 3471  This problem will not happen if the m-..
3471  of machine defines NO_UNION_TYPE.  of machine defines NO_UNION_TYPE.
3472    
3473    
3474  Copyright 1987,88,89,93,94,95,96,97,98,1999,2001,2002  Copyright 1987,88,89,93,94,95,96,97,98,1999,2001,2002,2004
3475    Free Software Foundation, Inc.    Free Software Foundation, Inc.
3476    
3477  Copying and redistribution of this file with or without modification  Copying and redistribution of this file with or without modification

Legend:
Removed from v.1.166  
changed lines
  Added in v.1.166.2.1

savannah-hackers-public@gnu.org
ViewVC Help
Powered by ViewVC 1.1.26