http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58145
--- Comment #3 from Hans-Peter Nilsson <hp at gcc dot gnu.org> --- (In reply to Jakub Jelinek from comment #2) > Created attachment 30653 [details] > gcc49-pr58145.patch > > Updated patch. Thank you very much, Jakub! The missing opportunity to learn trees :) is offset by far by the value of the promptness of the patch and the right generic incantations for the test-case! I'll test and regtest this, though the exact change in volatileness won't be tested beyond the test-case and the code-base where this was spotted. Still, apparently the patch can only add volatileness indicators where none was before, so should be safe even for other branches than trunk. It's obvious to you and other tree-ssa-savvy people, but IMHO its notable that fiddling with the test-case reveals that the bug seems limited to singleton-bit-field-structures of "natural" sizes; 8, 16, 32, (64 etc. where applicable). E.g. changing the bit-field-size to 9 or having two bitfields of 16 bits does not trig the bug.
