gcc released recently patch for this issue: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37179 Could You apply it to portage ? Patch affects stable versions of gcc so it is quite essential. Reproducible: Always Steps to Reproduce: 1.Buy Geode LX cpu 2.Try to compile mplayer on it (march=geode) 3.Try to run mplayer Actual Results: Illegal instruction Expected Results: mplayer should play the music ;) Just apply official patch and everything should be ok
Nice. Upstream has already applied this patch to branches.
According to the upstream bug, march=geode works fine. It only happens if you use -march=native. I just wanted to clear that up since the original post claims contrary to what upstream is saying.
(In reply to comment #2) > According to the upstream bug, march=geode works fine. It only happens if you > use -march=native. I just wanted to clear that up since the original post > claims contrary to what upstream is saying. > Then I don't get it, I have ffreep in my mplayer binary which was compiled with march=geode.
*** Bug 271343 has been marked as a duplicate of this bug. ***
(In reply to comment #4) > *** Bug 271343 has been marked as a duplicate of this bug. *** There stays "Falling back to -march=geode fixes the problem." when in fact -march=geode causes the problem. Will provide some detailed info about my specs soon.
As I said in http://bugs.gentoo.org/show_bug.cgi?id=274273 , mplayer overrides your CFLAGS. -march=geode therefore does not cause the problem, but the -march=native from mplayer causes it.
Um, this isn't fixed.
it is fixed(not even issue) since this was problem of mplayer, not gcc, USE flag custom-cflags for mplayer package solves the problem
(In reply to comment #8) > it is fixed(not even issue) since this was problem of mplayer, not gcc, USE > flag custom-cflags for mplayer package solves the problem > It does not change the fact that -march=native is broken for Geode CPUs. I have added these patches to the patchset for the next release