https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70240
--- Comment #5 from Jakub Jelinek <jakub at gcc dot gnu.org> --- (In reply to Eric Botcazou from comment #4) > > Better gimplify a and b before building the COND_EXPR. > > But then you undo the fix for PR middle-end/68215, so you make > c-c++-common/opaque-vector.c > regress again on a bunch of platforms. > > Besides gimplification is supposed to be fully recursive. Well, perhaps even better would be not to build all the trees, but use gimple_build () APIs that should also simplify stuff while building it on GIMPLE. But that might be too risky for stage4 now.
