http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45094
--- Comment #9 from Siarhei Siamashka <siarhei.siamashka at gmail dot com> 2010-12-18 15:43:39 UTC --- Can this bug get a "[4.5 regression]" header please? Even though the bug existed in gcc sources since 2007 (see the link in comment 1), the reported wrong-code problem itself was apparently latent until gcc 4.5, and is not reproducible with older gcc versions.