http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55545
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> 2012-11-30 13:11:38 UTC --- (In reply to comment #1) > Do you also have revision numbers in those logs? Unfortunately not, but I can see that rev. 193357 was definitely ok and rev. 193534 definitely broken. Maybe rev. 193429 also was still ok, but I'm not sure on that. Happens with both -O3 (base flags) and -O3 -funroll-loops -fpeel-loops (peak flags). In the larger revision range the only suspicious patches are the one for PR51447 and the gcse reg-pressure work from Bin Cheng. It seems to be compiling build-disjuncts.c btw.
