Sat 26 Apr 2008 05:18:10 PM UTC, comment #12:
Glad it compiles now!
|
Sat 26 Apr 2008 05:01:13 PM UTC, comment #11:
OK, please close this.
All builds OK now.
I am really sorry for the noise.
Nick
|
Sat 26 Apr 2008 01:47:51 PM UTC, comment #10:
Sigh :-/
After two days to build and get a working toolset with a later GCC:
gcc version 4.2.3
and tested on a lot of other code (Fluxbox etc.), I now get this error (cvs update today):
rc.cpp: In constructor 'gnash::RcInitFile::RcInitFile()':
rc.cpp:93: error: 'DEFAULT_STREAMS_TIMEOUT' was not declared in this scope
rc.cpp: In member function 'void gnash::RcInitFile::loadFiles()':
rc.cpp:129: error: 'SYSCONFDIR' was not declared in this scope
make[2]: *** [libgnashbase_la-rc.lo] Error 1
Sorry to be a pain...
Nick
|
Thu 24 Apr 2008 04:32:03 PM UTC, comment #9:
Benjamin,
I agree with you. Let me do some investigation on this issue.
Nick
|
Thu 24 Apr 2008 03:58:48 PM UTC, comment #8:
Right, we've cleared up that the failure isn't in Stage.cpp any more, so not to do with const_reverse_iterator.
A segfault in gcc is not Gnash's problem, whatever else the compiler can cope with. If there were any indication of what's causing it, we might try a workaround (though changing perfectly good code for buggy compilers is annoying). Without even that information, it would be silly for us to spend time working out bugs in a compiler (and an old one at that) rather than fixing Gnash.
|
Thu 24 Apr 2008 03:49:45 PM UTC, comment #7:
Nick, how about trying a very small testcase for
use of const_reverse_iterator over strings with that compiler ?
Seems worth to find out what fails exactly, if not else for
filing a bug report over GCC guys...
|
Thu 24 Apr 2008 03:45:11 PM UTC, comment #6:
Newer compiler? Well, not really. Last time I tried to upgrade it broke a lot of things.
But I don't understand this. I build Xorg git, KDE svn, Fluxbox git etc. etc. (in fact everything I build from source) and never ever had any issues like this.
Nick
|
Thu 24 Apr 2008 03:33:02 PM UTC, comment #5:
I don't think there's much doubt that that one's a gcc bug...
Can you use a newer compiler? There's only so much we can do to workaround compiler bugs, and with that one there's nothing at all to go on.
|
Thu 24 Apr 2008 03:27:24 PM UTC, comment #4:
Ok, benjamin - thanks.
GCC 3.4.6, Glibc 3.4.6.
OK, this seems to have fixed the build issue, but now it looks like GCC gets in an enternal loop and segfaults straight after - I have never seen this on my system here in years:
make[3]: Leaving directory `/home/hdb/cvs/gnash/server/asobj'
Making all in parser
make[3]: Entering directory `/home/hdb/cvs/gnash/server/parser'
if /bin/sh ../../libtool --tag=CXX --mode=compile g++ -DHAVE_CONFIG_H -I. -I. -I../.. -I../../libnet -I../../backend -I../../server -I../../server/swf -I../../server/parser -I../../server/vm -I../../server/asobj -I../../libbase -I../../libgeometry -I../../libmedia -I../../libamf -I../.. -I/usr/include/libxml2 -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -pthread -I/usr/local/include/boost-1_33_1 -I/usr/include/freetype2 -pthread -I/usr/include/gstreamer-0.10 -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -I/usr/include/libxml2 -I../../libmedia/gst -g -O2 -pthread -pthread -W -Wall -Wcast-align -Wcast-qual -Wpointer-arith -Wreturn-type -MT BitmapMovieDefinition.lo -MD -MP -MF ".deps/BitmapMovieDefinition.Tpo" -c -o BitmapMovieDefinition.lo BitmapMovieDefinition.cpp; \
then mv -f ".deps/BitmapMovieDefinition.Tpo" ".deps/BitmapMovieDefinition.Plo"; else rm -f ".deps/BitmapMovieDefinition.Tpo"; exit 1; fi
g++ -DHAVE_CONFIG_H -I. -I. -I../.. -I../../libnet -I../../backend -I../../server -I../../server/swf -I../../server/parser -I../../server/vm -I../../server/asobj -I../../libbase -I../../libgeometry -I../../libmedia -I../../libamf -I../.. -I/usr/include/libxml2 -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -pthread -I/usr/local/include/boost-1_33_1 -I/usr/include/freetype2 -pthread -I/usr/include/gstreamer-0.10 -I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include -I/usr/include/libxml2 -I../../libmedia/gst -g -O2 -pthread -pthread -W -Wall -Wcast-align -Wcast-qual -Wpointer-arith -Wreturn-type -MT BitmapMovieDefinition.lo -MD -MP -MF .deps/BitmapMovieDefinition.Tpo -c BitmapMovieDefinition.cpp -fPIC -DPIC -o .libs/BitmapMovieDefinition.o
g++: Internal error: Segmentation fault (program cc1plus)
Nick
|
Thu 24 Apr 2008 02:34:32 PM UTC, comment #3:
Well, once I'd fixed another compilation bug on win32, this one cropped up on our build farm windows 2000 machine. It seems I could fix that one as well, so give it another try.
|
Thu 24 Apr 2008 01:01:10 PM UTC, comment #2:
What compiler are you using? The error line is this:
for (std::string::const_reverse_iterator it = _alignMode.rbegin();
it != _alignMode.rend();
++it)
{ /* ... */ }
which seems fine to me, and compiles on all our build farm machines.
|
Thu 24 Apr 2008 11:40:29 AM UTC, comment #1:
Actually, a bum steer. It just so happened that the boost warnings happened exactly the same place as build failure, so I was lead astray:
Stage.cpp: In member function `void gnash::Stage::setAlignMode(const std::string&)':
Stage.cpp:208: error: no match for 'operator!=' in 'it != std::basic_string<_CharT, _Traits, _Alloc>::rend() [with _CharT = char, _Traits = std::char_traits<char>, _Alloc = std::allocator<char>]()'
make[3]: *** [Stage.lo] Error 1
I don't really understand C++ much...
Nick
|
Thu 24 Apr 2008 11:29:03 AM UTC, original submission:
Hi all,
Recently being able to use gnash again, latest CVS build fails. Although this is not an gnash issue, I am recording here in case other people get the issue (as gnash builds fails due to it).
Analyse shows this is a boost issue that causes the failure, ref:
http://lists.boost.org/boost-users/2006/04/19170.php
Patch here:
http://lists.boost.org/boost-users/att-19170/implementation_help.hpp.const_cast_0.patch
Nick
|