AbiWord crashes for me when I use -fomit-frame-pointers. Actually, it crashes when I try to copy/cut/paste anything, even (I think) DND. The AbiWord developers refuse to help me -- it's the whole Heisenburg thing -- how do you debug with -fomit-frame-pointers turned on? Filtering fixes the problem.
Steps to Reproduce:
1. Install AbiWord on x86 Linux with CFLAGS="-march=athlon-xp -Os -fomit-frame-pointer"
2. Run AbiWord and try to cut/copy/paste something. Anything.
4. recompile without -fomit-frame-pointer
5. works fine
Portage 2.0.50-r10 (default-x86-1.4, gcc-3.3.4, glibc-22.214.171.12440420-r1,
System uname: 2.6.9-rc1-mm1 i686 AMD Athlon(tm) XP 2100+
Gentoo Base System version 1.4.16
CFLAGS="-march=athlon-xp -Os -fomit-frame-pointer"
CONFIG_PROTECT="/etc /usr/X11R6/lib/X11/xkb /usr/kde/2/share/config
/usr/kde/3/share/config /usr/lib/mozilla/defaults/pref /usr/share/config
CONFIG_PROTECT_MASK="/etc/gconf /etc/terminfo /etc/env.d"
CXXFLAGS="-march=athlon-xp -Os -fomit-frame-pointer"
FEATURES="autoaddcvs ccache sandbox userpriv"
USE="3dnow 3ds S3TC X X509 Xaw3d aalib acpi acpi4linux alsa apache1 apache2
async avi berkdb bonobo caps cddb cdr cdrom chroot crypt cups curl dga dnd dvd
editor encode ethereal evms2 ext-png ext-zlib fam fbcon fbdev flac flash
foomaticdb freetype gcj gdbm ggi gif gnome gnomedb gphoto2 gpm gtk gtk2 guile
libg++ libgda libwww mad maildir md5sum mdb mikmod ming mmx mng monkey
moznocompose moznoirc moznomail moznoxft mpeg mpeg4 mplayer msn music nas
ncurses net nowin nvidia offensive ofx oggvorbis openal opengl openssh pam
pdflib perl plotutils png pnp ppds psyco python qhull quicktime radeon readline
ruby samba sdl slang speex spell sse ssl tcltk tcpd tiff truetype usb videos
vim-with-x wmf wxwindows x86 xface xine xml xml2 xmms xosd xv xvid yahoo zlib"
Created attachment 38554 [details, diff]
patch to abiword-2.0.10.ebuild
filters -fomit-frame-pointer from AbiWord, thus fixing an issue of CRASH on
Created attachment 38555 [details, diff]
patch to abiword-2.0.10.ebuild
oops, above patch is reversed. This fixes it.
i need more reports to put such a restriction in.
Shouldn't this be up to the user, since those flags cause problems with debugging on most packages.
Yes, those flags do cause problems with debugging. They also cause the program to crash for me. Hence they are removed.
I can duplicate the crash on two separate machines. If anyone wants to try, go ahead -- try my CFLAGS.
I would love to have this be more specific -- maybe -fomit-frame-pointer doesn't get along with -Os? Still, I'd rather it have a _very_minor_ optimization removed so that it works for everyone, and then add it back in if it's really needed.
sorry,seems I didn't finish my thought :), I compiled with -fomit-frame-pointer here to test and didn't have that proplem. If it ends up being a combination of different cflags, then it is sadly the users issue.
I'll try a bit to find that combo, then.
How hard is it to filter a combo from cflags? Why does it become the user's issue -- weren't my cflags sane?
I'll go away if someone gives me per-package cflag settings and I can post big warnings in the forum about this issue.
Just to check...are you sure it's not the size-optimization (-Os) that's doing it? Does it get fixed if you use CFLAGS such as "-O2 -march=athlon-xp -fomit-frame-pointer"? And considering that the most recent 2.0 release is 2.0.14 (latest in Portage is 2.0.11), does upgrading help?
We're not going to filter all possible flag combinations, unless we can get multiple reports from a flag causing known breakage. Your CFLAGS are sane - try with i686 and -O2 maybe - but this doesn't mean we need to filter omit-fp for everyone.