| Summary: | [4.5] dev-libs/libisofs fails to compile on ARMv5 | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Markus Meier <maekke> |
| Component: | Current packages | Assignee: | Daniel Pielmeier <billie> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | arm, media-optical, toolchain |
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| URL: | http://gcc.gnu.org/PR45726 | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Bug Depends on: | 348898 | ||
| Bug Blocks: | |||
| Attachments: | dev-libs:libisofs-0.6.40:20101218-164738.log | ||
|
Description
Markus Meier
2010-12-20 08:25:23 UTC
Created attachment 257601 [details]
dev-libs:libisofs-0.6.40:20101218-164738.log
I guess the new versions of libburn and libisoburn compile? (In reply to comment #2) > I guess the new versions of libburn and libisoburn compile? dev-libs/libburn-0.9.0 yes, dev-libs/libisoburn-0.6.6 needs libisofs, which fails... I have contacted upstream about this issue and will add his remarks here. Although the error occurs in libisofs/messages.c the reported line number seems to be from the generated assembler code. Somehow we need to identify the code in messages.c which generates the failing assembler code. Maybe the toolchain guys can help us out here. Also can you try if it builds with debugging enabled and without -02 optimisations. (In reply to comment #4) > Also can you try if it builds with debugging enabled and without -02 > optimisations. CFLAGS="-O0 -march=armv5te -pipe" CXXFLAGS="${CFLAGS}" USE="debug" emerge -1 libisofs emerged w/o issues like that. fails the same way with -O1 as with -O2 (debug USE-flag doesn't matter). (In reply to comment #6) > fails the same way with -O1 as with -O2 (debug USE-flag doesn't matter). What about -O3? Can you try it as well? (In reply to comment #7) > (In reply to comment #6) > > fails the same way with -O1 as with -O2 (debug USE-flag doesn't matter). > > What about -O3? Can you try it as well? fails as well (+/-debug doesn't change anything) please ignore my comment #0 as libisofs-0.6.38 fails the same way with gcc-4.5.1. this looks like http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45726 - I'm rebuilding gcc with these patches. (In reply to comment #9) > this looks like http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45726 - I'm > rebuilding gcc with these patches. applying the patch from the bug solves the issue with libisofs (and the testcase from the gcc bug). @toolchain: it would be great if you could integrate it in our gcc patchset - thanks in advance. I'll include it in the 4.5.2 bump next week. Fixed in 4.5.2. |