media-gfx/graphviz-1.8.10 ebuild fails to compile on a VIA C3 800MHz (mini-itx system) CHOST="i586-pc-linux-gnu" CFLAGS="-march=i586 -m3dnow -O3 -pipe -fomit-frame-pointer" CXXFLAGS="${CFLAGS}" beambox root # emerge world Calculating world dependencies ...done! >>> emerge (1 of 1) media-gfx/graphviz-1.8.10 to / >>> md5 ;-) graphviz-1.8.10.tar.gz ... /bin/sh ../../libtool --mode=compile gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../../tools/vmalloc -I../../tools/sfio -I../../tools/ast -I../../cdt -march=i586 -m3dnow -O3 -pipe -fomit-frame-pointer -Wall -Wno-unknown-pragmas -c exeval.c gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../../tools/vmalloc -I../../tools/sfio -I../../tools/ast -I../../cdt -march=i586 -m3dnow -O3 -pipe -fomit-frame-pointer -Wall -Wno-unknown-pragmas -c exeval.c -fPIC -DPIC -o exeval.lo exeval.c: In function `eval': exeval.c:1225: Internal compiler error in delete_insn, at cfgrtl.c:139 Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions. make[3]: *** [exeval.lo] Error 1 make[3]: Leaving directory `/var/tmp/portage/graphviz-1.8.10/work/graphviz-1.8.10/tools/expr' make[2]: *** [all-recursive] Error 1 make[2]: Leaving directory `/var/tmp/portage/graphviz-1.8.10/work/graphviz-1.8.10/tools' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory `/var/tmp/portage/graphviz-1.8.10/work/graphviz-1.8.10' make: *** [all-recursive-am] Error 2 !!! ERROR: media-gfx/graphviz-1.8.10 failed. !!! Function src_compile, Line -143, Exitcode 2 !!! (no error message)
Hi Justin Thank you for the bug report! The: >exeval.c:1225: Internal compiler error in delete_insn, at cfgrtl.c:139 makes me suspect bad hardware as the #1 thing. Could you please: 1. confirm that you do not have a bad memory issue (running memtest86 is the best one), overclocking and did not experience such faults before? 2. Could you please a) add "noclean" to FEATURES in your make.conf b) emerge this package again (via emerge graphviz, not world!) and when it breaks (supposedly at the same place) cd /var/tmp/portage/graphviz-1.8.10/work/${P} and run "make" manually. Does it immediately stop with the same error message or does it continue to build for a while and the breaks again at some new point? If you can confirm that this is not a hardware issue, please supply some additional information: 1. your profile, please make a notice if you do not have cuurent versions of glibc/gcc installed. 2. Do you use stable profile or routinely --update world with "~x86" set on? George
1. memtest86 (v3.0) result: All memory ok. (Damn those memory tests take a while with 512MB of RAM...) 2. Build halts in same spot. # cd /var/tmp/portage/graphviz-1.8.10/work/graphviz-1.8.10/ # make This builds for a little while (a few pages of text), then bombs with Making all in expr make[3]: Entering directory `/var/tmp/portage/graphviz-1.8.10/work/graphviz-1.8.10/tools/expr' /bin/sh ../../libtool --mode=compile gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../../tools/vmalloc -I../../tools/sfio -I../../tools/ast -I../../cdt -march=i586 -m3dnow -O3 -pipe -fomit-frame-pointer -Wall -Wno-unknown-pragmas -c exeval.c gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../../tools/vmalloc -I../../tools/sfio -I../../tools/ast -I../../cdt -march=i586 -m3dnow -O3 -pipe -fomit-frame-pointer -Wall -Wno-unknown-pragmas -c exeval.c -fPIC -DPIC -o exeval.lo exeval.c: In function `eval': exeval.c:1225: Internal compiler error in delete_insn, at cfgrtl.c:139 Please submit a full bug report, with preprocessed source if appropriate. See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions. make[3]: *** [exeval.lo] Error 1 make[3]: Leaving directory `/var/tmp/portage/graphviz-1.8.10/work/graphviz-1.8.10/tools/expr' make[2]: *** [all-recursive] Error 1 make[2]: Leaving directory `/var/tmp/portage/graphviz-1.8.10/work/graphviz-1.8.10/tools' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory `/var/tmp/portage/graphviz-1.8.10/work/graphviz-1.8.10' make: *** [all-recursive-am] Error 2 A repeat of this builds for a short time with the exact same text, then crashes in the same spot. 3. I use a stable profile with the latest glibc/gcc. I have never used "~x86". -Justin
Hi Justin Thanks for those checks, now time fore more questions ;). >I use a stable profile with the latest glibc/gcc. I have never used "~x86". But is this a default-1.0 or default-x86-1.4? Basically is this a gcc-3.2 or gcc-2.95.3 system? Also, could you please include the output of qpkg -i libpng? I am looking at what versions of libpng you have installed at the moment. This package seems to have an issue with libpng-1.2 (so it depends on =libpng-1.0*), though the issue may be broader than what I know about it as of now. George
Hi Justin. another quick question: do you by any chance have LC_COLLATE set to anything? Could you please check what echo $LC_COLLATE returns or if this var is defined anywhere on your system? George
root # echo $LC_COLLATE root # qpkg -i libpng media-libs/libpng-1.2.4 * libpng [ http://www.libpng.org/ ] media-libs/libpng-1.0.12-r1 * libpng [ http://www.libpng.org/ ] media-libs/libpng-1.2.4 libpng [ http://www.libpng.org/ ] media-libs/libpng-1.0.12-r1 libpng [ http://www.libpng.org/ ] media-libs/libpng-1.2.5 Portable Network Graphics library [ http://www.libpng.org/ ]
Hi Justin. Your problem does not seem to be directly related, but you still might want to give it a try: In an attempt to get rid of libpng-1.0, which does not want to peacefully coexist with 1.2 and is causing niticeable trouble I modified the graphviz ebuild and made it compile without libpng support (graphviz is the last app in the portage that requires libpng-1.0). The ebuild is media-gfx/graphviz-1.8.10-r1, it is masked by both KEYWORDS and in package mask. You can try unmasking and building it. Please report the result if you do. George
With the latest Portage (2.0.45-r5) and GCC (3.2.1), media-gfx/graphviz-1.8.10 ebuild succeeds on my VIA C3 800MHz (mini-itx system). Happily enough, this bug has been fixed.