According to amodra's analysis, "-mno-sched-prolog breaks function parameter debug location lists" on powerpc. http://sourceware.org/bugzilla/show_bug.cgi?id=10231 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40473 I have not yet tried to reproduce this with gcc-vta/rawhide yet, but it impacts F11 for sure, so a backport of the patch would be appreciated.
First of all, the kernel shouldn't be using such options. The comment in kernel's Makefile talks about working around some gcc bug with -fno-omit-frame-pointer, has that bug even been filed to gcc?
(In reply to comment #1) > First of all, the kernel shouldn't be using such options. [...] Maybe, but if the patch now committed to gcc mainline can be backported, we can fight that battle another time.
Fixed in gcc-4.4.1-16.