| Summary: | doxygen-1.3.8 compile fails @ pngwtran.c with Assembler message (K6-II, ) | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Ervin Peters <ervin.peters> |
| Component: | Current packages | Assignee: | Steve Arnold <nerdboy> |
| Status: | RESOLVED UPSTREAM | ||
| Severity: | minor | ||
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | x86 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
I forget to mention: root@redmagic root # gcc --version gcc (GCC) 3.3.4 20040623 (Gentoo Linux 3.3.4-r1, ssp-3.3.2-2, pie-8.7.6) Copyright I forget to mention: root@redmagic root # gcc --version gcc (GCC) 3.3.4 20040623 (Gentoo Linux 3.3.4-r1, ssp-3.3.2-2, pie-8.7.6) Copyright © 2003 Free Software Foundation, Inc. ervin Have you tried -O2 with -fomit-frame-pointer? In the past, -O3 has been somewhat problematic; I usually stick with -O2 myself (haven't used a k6 in quite some time). I guess it would be good to know exactly which flags to filter... http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11689 shows it is a gcc error that should have been fixed: > Thus, the gcc compiler generates unsuitable code for the assembler. > The offending assembly lines contain a .loop statement with a branch target > that seems likely to be too large. One workaround mentioned ist: > The workaround for 3.3 is to use -fno-branch-count-reg. I cannot test it immediate, because my K6-II/550 is updating --deep world - and this will take a while (2 of 128 app-office/openoffice-ximian..). ervin Cool, if you could narrow it down to that flag, we can filter it. Also, I've been using 3.4.3 without any problems of note, so if it's already fixed upstream then they won't need another bug... Closing; re-open if you still have an issue. |
some output while emerging doxygen [..] gcc -c -pipe -Wall -W -fno-exceptions -O3 -march=k6-2 -funroll-loops -fomit-frame-pointer -pipe -o ../objects/pngwrite.o pngwrite.c gcc -c -pipe -Wall -W -fno-exceptions -O3 -march=k6-2 -funroll-loops -fomit-frame-pointer -pipe -o ../objects/pngwtran.o pngwtran.c {standard input}: Assembler messages: {standard input}:11395: Error: value of ffffffffffffff7b too large for field of 1 bytes at 0000000000008996 gmake[2]: *** [../objects/pngrtran.o] Fehler 1 gmake[2]: *** Warte auf noch nicht beendete Prozesse... gmake[2]: Leaving directory `/var/tmp/portage/doxygen-1.3.8/work/doxygen-1.3.8/libpng' [..] my make.conf values: root@redmagic peters # grep ^[^#] /etc/make.conf USE="-3dfx 3dnow aalib acl -afs -adns alsa apache2 arts avi berkdb -bidi bonobo -canna cdr cups dga directfb doc encode fastcgi kbcon flash foomaticdb -freewnn gb gd gdbm ggi gnome gps gtk gtk2 gtkhtml imap imagemagik imlib -java jpeg ldap libg++ libgda mad mcal mikmod mmx mozilla mpeg ncurses nls nocardbus -oci8 oggvorbis opengl pam -pcmcia -pda perl plotutils png postgres readline samba sasl scanner sdl slang slp -sse ssl svga tetex tiff truetype unicode -usb wmf X Xaw3d xml xml2 xmms xv yaz zlib x86" CHOST="i586-pc-linux-gnu" CFLAGS="-O3 -march=k6-2 -funroll-loops -fomit-frame-pointer -pipe" CXXFLAGS="${CFLAGS}" PORTDIR_OVERLAY="/usr/local/portage" MAKEOPTS="-j2" AUTOCLEAN="yes" [..] It seems it is a problem related to the 'march=K6-2' setting. On pentium I,II,III CPU it compiles without any problem, on an k6-III same problem. Searching bugs around about libpng (pngtrans.c resides in a subdir named libpng) shows this: http://bugs.gentoo.org/show_bug.cgi?id=50931 or better: http://bugs.gentoo.org/show_bug.cgi?id=12561 Setting march=i586 or removing -fomit-frame-pointer avoids this error, it seems to be a gcc/binutils related problem. ervin Reproducible: Always Steps to Reproduce: 1. emerge doxygen with k6 march and -fomit-frame-pointer Setting 2. 3.