https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117599
--- Comment #7 from Xi Ruoyao <xry111 at gcc dot gnu.org> --- (In reply to chenglulu from comment #6) > (In reply to Xi Ruoyao from comment #4) > > We have things like > > > > # VUSE <.MEM_74> > > _48 = VIEW_CONVERT_EXPR<unsigned long[10]>(b575[i_6])[8]; > > ./gcc/testsuite/gcc.dg/torture/bitint-12.c:39:1: error: type mismatch in > > 'array_ref' > > <unnamed-signed:63> > > > > where b575[i_6] is _BitInt(575). On x86 we have unsigned long[9] instead of > > unsigned long[10], I guess the difference comes from we must align _BitInt > > to 16B boundary? > > Before this patch, Yang Yujie submitted several patches that modified the > public layer code, so maybe you need to update the code? I have r12-3265 here and I cannot see any commit from Yujie to tree-cfg.cc (where the GIMPLE verifier stays). Adding -fno-checking allows the test to pass so maybe in your CI you've used something like --enable-checking=release and happened to hide the issue?
