Jeffrey Bridge
2002-Jul-19 19:59 UTC
[vorbis-dev] compile optimization error in libvorbis-1.0
Wow, I've never seen an FSF gcc crap out like this on a non-overheating system. When lib/envelope.c is compiled with -march=i686, gcc gets a signal 11, but not with -march=i586. I'm not sure whether to blame gcc or libvorbis. There are apparently three other files in libvorbis where this same behavior is seen. jeffrey@diddl:/usr/src/build/libvorbis-1.0/lib$ gcc -DPACKAGE=\"libvorbis\" -DVERSION=\"1.0\" -DHAVE_DLFCN_H=1 -DHAVE_ALLOCA_H=1 -DHAVE_ALLOCA=1 -DHAVE_SQRTF=1 -DHAVE_LOGF=1 -DHAVE_EXPF=1 -DHAVE_ACOSF=1 -DHAVE_ATANF=1 -DHAVE_FREXPF=1 -DHAVE_RINTF=1 -DHAVE_FLOORF=1 -I. -I. -I../include -I/usr/include -ffast-math -mno-ieee-fp -D_REENTRANT -fsigned-char -g -O2 -march=i686 -DUSE_MEMORY_H -c envelope.c -fPIC -DPIC -o .libs/envelope.lo gcc: Internal compiler error: program cc1 got fatal signal 11 Replacing i686 with i586 makes it work. This is with glibc 2.2.5, gcc 2.95.3 release. ~jeffrey :j --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
Michael Smith
2002-Jul-19 20:34 UTC
[vorbis-dev] compile optimization error in libvorbis-1.0
At 09:59 PM 7/19/02 -0500, you wrote:>Wow, I've never seen an FSF gcc crap out like this on a non-overheating >system. When lib/envelope.c is compiled with -march=i686, gcc gets a >signal 11, but not with -march=i586. I'm not sure whether to blame gcc >or libvorbis. There are apparently three other files in libvorbis where >this same behavior is seen.Very well known problem. Bug in gcc 2.95.x, gets triggered by i686 optimisation in combination with some other flags that libvorbis uses. Michael <p>--- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
vorbis@fuchs.offl.uni-jena.de
2002-Jul-20 01:41 UTC
[vorbis-dev] compile optimization error in libvorbis-1.0
On Fri, Jul 19, 2002 at 09:59:23PM -0500, Jeffrey Bridge wrote:> Wow, I've never seen an FSF gcc crap out like this on a non-overheating > system. When lib/envelope.c is compiled with -march=i686, gcc gets a > signal 11, but not with -march=i586. I'm not sure whether to blame gcc > or libvorbis. There are apparently three other files in libvorbis where > this same behavior is seen. > > jeffrey@diddl:/usr/src/build/libvorbis-1.0/lib$ gcc > -DPACKAGE=\"libvorbis\" -DVERSION=\"1.0\" -DHAVE_DLFCN_H=1 > -DHAVE_ALLOCA_H=1 -DHAVE_ALLOCA=1 -DHAVE_SQRTF=1 -DHAVE_LOGF=1 > -DHAVE_EXPF=1 -DHAVE_ACOSF=1 -DHAVE_ATANF=1 -DHAVE_FREXPF=1 > -DHAVE_RINTF=1 -DHAVE_FLOORF=1 -I. -I. -I../include -I/usr/include > -ffast-math -mno-ieee-fp -D_REENTRANT -fsigned-char -g -O2 -march=i686 > -DUSE_MEMORY_H -c envelope.c -fPIC -DPIC -o .libs/envelope.lo > gcc: Internal compiler error: program cc1 got fatal signal 11 > > Replacing i686 with i586 makes it work. This is with glibc 2.2.5, gcc > 2.95.3 release. >Can't be a problem of Vorbis. Even the worst C code should not generate a SIGSEGV. I've seen thius bug also with gcc-2.95.3. But there i686 generated a SIGSEGV, i586 worked. -- Frank Klemm <p>--- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.