------- Comment #36 from amylaar at gcc dot gnu dot org 2007-02-06 11:17 ------- (In reply to comment #34) > I'll prepare a patch to revert my 2004 change too.
I suspect that a 100% literal reversion will run into problems where the use of a global variable will result in the the analysis of one function being applied to another function. It will probably work (modulo minor merge conflicts) if instead of bringing the global variable current_function_nothrow back, you define a nothrow bit in struct function and #define current_function_nothrow cfun->nothrow . -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29487