Summary: | [ICE/4.6] media-libs/x264-0.0.20120707 on armv7a-hardfloat - encoder/ratecontrol.c:1794:1: error: unable to find a register to spill in class 'VFP_LO_REGS' | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | jannis <kripton> |
Component: | [OLD] Development | Assignee: | Gentoo Toolchain Maintainers <toolchain> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | esigra, tomwij, zerochaos |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | ARM | ||
OS: | Linux | ||
URL: | http://gcc.gnu.org/PR48808 | ||
See Also: | http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48808 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
build.log
emerge --info |
Description
jannis
2013-01-18 07:36:00 UTC
Created attachment 336008 [details]
build.log
Created attachment 336010 [details]
emerge --info
x264-devs say it's gcc's fault since it's failing in native C code, no asm involved. gcc devs say it might be related to: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49422 or http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48808 and that I should try out vanilla gcc or a more recent version (than gcc 4.6.3) I'm just compiling gcc-4.6.3 with the patches from http://gcc.gnu.org/viewcvs?view=revision&revision=174540 mentioned in http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48808 I'll know more in about 2hrs Okay, that took a little longer. I patched the two "reload*.c" files from the links posted, emerged gcc-4.6.3 and x264 compiles fine. I encoded one video file to test the software and the result (encoded file) plays back well. Compiling with gcc-4.5.4 also compiles and works fine so this seems to be a bug of gcc-4.6.3 on arm (and it should thus not be stabilized there). Maybe I'll test 4.6.1 or 4.6.2 later. Or the toolchain-devs add the patch to the gcc's patchset. This also occurs when attempting to build icedtea:6 with gcc 4.6.3 this doesn't block gcc 4.6 stable request, it's not even keyworded for arm, let alone stable. this also causes failures on icedtea:7, same failures. thanks for tracking down the PR. does the patch in the PR also fix the icedtea failure? Working on figuring that out now Ryan. Rick says that it doesn't fix it there, I'm doing a fresh chroot now with the patches. (In reply to comment #10) > Working on figuring that out now Ryan. Rick says that it doesn't fix it > there, I'm doing a fresh chroot now with the patches. My main problem currently is that user_patch doesn't seem to pick up patches named blah.c. I'm rebuilding gcc again, with the patches actually applied this time. Please give me a few and I'll report back. This fixes the ICE for x264, icedtea:6 and icedtea:7. Sadly icedtea:6 and icedtea:7 still don't actually install, but that is a significantly different issue. The attached patches should be rolled into our 4.6.3 patchset asap. Fixed in 4.6.3 p1.11. |