http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53508
Bug #: 53508 Summary: function.c:stack_protect_epilogue call to predict_insn_def with wrong direction of prediction Classification: Unclassified Product: gcc Version: 4.8.0 URL: http://gcc.gnu.org/ml/gcc-patches/2012-05/msg01361.htm l Status: UNCONFIRMED Severity: normal Priority: P3 Component: rtl-optimization AssignedTo: unassig...@gcc.gnu.org ReportedBy: h...@gcc.gnu.org The message at the URL and the (original) summary says it all; the branch is *around* a noreturn call, so the call to predict_insn_def has the prediction direction inverted. The doc wrong too; corrected with revision 187934 on trunk. The buglet is as far back as the 4.3-branch. The generated code has the call moved to a "cold" position, so the generated code does not seem to be affected in the trivial case.