It seems gcc-3.3.4 doesn't work with the current altivec patch for 3.3.2... if [ "${ARCH}" = "ppc" -o "${ARCH}" = "ppc64" ] then epatch ${FILESDIR}/3.3.2/gcc332-altivec-fix.patch fi I am just trying to get a released stable gcc that will compile QT... trying it now... You can even assign this to me and Ill return the results building from stage1 a ppc32 system today. PPC64 note, the patch affects ppc64 too...
stage1 compiles and run qt 3.3.2 compiles Things I did: needed to install the latest gcc-config-1.3.6 changed the keyword so that bootstrap thinks it is stable. commented out the whole altivec patch for gcc 3.3.2 Keyworded gcc-3.3.4 has ppc
if you have an altivec capable cpu please make sure that altivec code builds fine. I'll have a look after my exams
Compiling mplayer with altivec should be a good test (check if you can watch movies properly - what exactly was the problem Lu?) ...
Doing that JUST now ... will have it playing tomorrow ...
The vip bugs currently related to gcc on ppc are: 1 altivec misoptimized on gcc-3.3.3 (mainline, the hammer branch addresses to that one) [testcase:: ffmpeg mpeg4 codec] 2 Optimization related issues with gcc-3.3.3 (every) [testcase:: diffutils] Both problem are solved on gcc-3.4.0 and I'm chrash-testing on most of the code I could build, so far gcc-3.4.0 was ok if you don't have wacky C++ code or label before branch situation. If gcc-3.3.4 addresses the problems I described before we could mark it ppc after some automated test.
recompiled diffutils, mplayer and ffmpeg mplayer is usable plays movies and all diff -Naur somwthing something.orig returns beliavable results xine works fine and plays movies haven't yet found a nice way to test ffmpeg... someone can give me something that really breaks things like a 1 2 3 boom procedure and Ill do it... Trying to build atlas and lapack-atlas... Now
dev-libs/atlas-3.6.0 compiled and is used by R and lush
ppc64 has gcc-3.4.x marked stable now which solves the problem. ppc is using a non standard toolchain for the 3.3.x so reassigning bug to the ppc team
gcc-3.4.1-r2 will be marked stable within the week
3.4 stable on ppc, we can still revisit gcc-3.3.4 later on if needed. tchiwam, your call :-)
I'm going to close this one out. in the ppc64 context, the 3.3 series of gcc isn't something we're making use of anymore.
