Fri 18 Oct 2013 04:14:34 PM UTC, comment #18:
I have not had a chance to install dietlibc and build RCS against that, but wonder anyway if the crash still manifests with RCS 5.9.1 (released 2013-10-04). Any news?
|
Sat 21 Aug 2010 11:39:10 AM UTC, comment #17:
No worries. I will mark this bug postponed for now. When i am able to reproduce the environment, i will continue then.
|
Thu 19 Aug 2010 06:21:15 PM UTC, comment #16:
I'm sorry for the late reply. rcs segfaults and the test ends... Sorry I didn't capture stderr, nore did I get a BT.
You can duplicate my environment exactly. This is Debian with the dietlibc-dev packages. The install/VM is from ThoughtPolice.
Setting up a dietlibc build environment should be simple wherever you are.
|
Sun 15 Aug 2010 12:28:45 PM UTC, comment #15:
Thanks for the _POSIX_SYMLOOP_MAX patch; i have applied it.
Thanks also for collecting the other information. I have a doubt, however:
Are you sure you did "make check DEBUG=1"? I expect to see lots of output (from the script doing set -x), but all i see in the log file is:
#ci+rcs -L failed
FAIL: alive.test
|
Mon 02 Aug 2010 06:07:46 PM UTC, comment #14:
diff -u -r1.1 ./src/rcsedit.c
--- ./src/rcsedit.c 2010/08/02 16:42:51 1.1
+++ ./src/rcsedit.c 2010/08/02 16:43:16
@@ -48,6 +48,10 @@
#include "b-isr.h"
#include "b-kwxout.h"
+#ifndef _POSIX_SYMLOOP_MAX
+#define _POSIX_SYMLOOP_MAX 8
+#endif
+
struct editstuff
{
struct fro *fedit;
Also note the files attached.
|
Thu 22 Jul 2010 08:04:47 AM UTC, comment #13:
Ping (to move things forward).
Could you please build/test 5.7.92 and post:
- config.log
- output of "make"
- output of "make check DEBUG=1"
- (assuming "make check" fails) all src/a.* files
I'd like to either get more info on the missing-date problem, or close this bug in the next week or two.
|
Tue 13 Jul 2010 03:31:25 PM UTC, comment #12:
It only relates the the Dietlibc build, not each specific issue with building on Dietlibc.
Perhaps child tickets should be created to track each issue. I'll test this.
For the missing date problem I've been told that some functionality(like _Exit) was removed/"never added" from the library to make it smaller. Thus there is likely a warning that the date produced by function X will always be empty in the build log.
|
Tue 13 Jul 2010 10:42:51 AM UTC, comment #11:
Support for missing _Exit has now been added to RCS and another pretest release 5.7.92 is now available (alpha.gnu.org). Please try that one.
(I am still unsure how this relates to the missing date problem, however.)
|
Mon 12 Jul 2010 11:05:32 AM UTC, comment #10:
OK, i see from:
http://www.fefe.de/dietlibc/exports
that dietlibc does not have _Exit at all.
|
Mon 12 Jul 2010 10:01:54 AM UTC, comment #9:
Thank you for building / testing 5.7.91.
Reviewing the comments so far, it seems like there is some discussion happening between comments #7 and #8 that i'm not privy to, the gist of which is that dietlibc does not provide _Exit (without some feature-test macros defined) -- is that correct?
Could you please post config.log and the output of "make check"?
Lastly, with _Exit properly defined, does the original problem (missing date information in the RCS file) still occur?
|
Thu 08 Jul 2010 03:46:32 PM UTC, comment #8:
Yes, it's true I don't get features.h at all. It works the other way around, that doesn't seam helpful for me, I don't see how one could force new features to exist.
|
Thu 08 Jul 2010 03:43:50 PM UTC, comment #7:
http://www.kernel.org/doc/man-pages/online/pages/man2/_exit.2.html
I'm not sure if my build environment is broken, but it looks like this _Exit VS _exit issue can be solved with this:
#include <features.h>
#nif _XOPEN_SOURCE >= 600 || _ISOC99_SOURCE
#define _Exit _exit
#endif
..or something like that.
|
Wed 07 Jul 2010 03:52:50 PM UTC, comment #6:
0x0804b8e1 in lock_on (delta=0x8066200) at b-excwho.c:240
240 for (struct link *ls = GROK (locks); ls; ls = ls->next)
#1 0x0804b97b in addlock (delta=0x8066200, verbose=true) at b-excwho.c:268
|
Wed 07 Jul 2010 03:47:30 PM UTC, comment #5:
mmestnik@build-dietlibc:~$ gdb --args ~/src/diet/rcs-official/rcs-5.7.91/src/ci -l test
GNU gdb 6.8-debian
Copyright (C) 2008 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "i486-linux-gnu"...
(no debugging symbols found)
(gdb) r
Starting program: /home/mmestnik/src/diet/rcs-official/rcs-5.7.91/src/ci -l test
(no debugging symbols found)
test,v <-- test
Program received signal SIGSEGV, Segmentation fault.
0x0804b8e1 in lock_on ()
(gdb) bt
#0 0x0804b8e1 in lock_on ()
#1 0x0804b97b in addlock ()
#2 0x08049d0d in main ()
(gdb)
I'll try building with debug.
|
Wed 07 Jul 2010 02:29:05 PM UTC, comment #4:
This would be based after this download.
rcs-5.7.tar.gz 16-Jun-1995 03:00 276K
-rw-r--r-- 1 mmestnik mmestnik 276K 2004-05-11 11:45 rcs_5.7.orig.tar.gz
md5 8fd09ea9654cda128f8d5c337d3b8de7
sha1 20f9fc9b2e773a19f67a6ff95ca4147b2ee1667c
The 25 indicates this is the 25th build or time Ubuntu has modified the source. Fixing bugs in the package or the program, likely applying security or back-ported features.
I'd try with that, once I find where I can download it.
|
Wed 07 Jul 2010 12:09:02 AM UTC, comment #3:
I don't know what that source is in relation to the 5.7.91 (pretest for 5.8) tarball. If you are willing to try building and testing 5.7.91, we can continue this bug. Otherwise, i would redirect you to the Ubuntu bug tracking system and close this bug after a week or so.
|
Tue 06 Jul 2010 08:07:57 PM UTC, comment #2:
Important:
This was rcs source as provided by Ubuntu:
rcs (5.7-25) unstable; urgency=low
* The "dogs that love the rain and chasing trains" release.
* debian/control:
+ Set section to 'vcs' to follow override.
+ Set Standards-Version to 3.8.1.
* debian/rules:
+ Use `dh_prep' instead of `dh_clean -k'.
+ Don't number exported patches.
* Drop Debian change added in rcs 5.7-13 (Nov 1999) which modifies the
first line of pic's output. As far as I can tell it's fine as-is, and
the change results in an extraneous line at the top of the page
(closes: #525291).
|
Tue 06 Jul 2010 07:35:21 PM UTC, comment #1:
Submitted by: mmestnik
|
Tue 06 Jul 2010 07:31:36 PM UTC, original submission:
I'll need to do some more testing, but the error I get is...
ci: /etc/init.d/nrpe,v:10: missing date
The version file has several spaces instead of a date. Depending on how this string is created/emitted this might need just a little adjustment to make a usable Dietlibc build.
A worthy effort would be to work toured removing all the warnings produced by the Dietlibc build to optimize the result.
Here is the resulting ,v file after committing a file of length 0:
head 1.1;
access;
symbols;
locks
root:1.1; strict;
comment @# @;
1.1
date ; author root; state Exp;
branches;
next ;
desc
@The description.
@
1.1
log
@Initial revision
@
text
@@
|