https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117423
--- Comment #20 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Sam James from comment #17) > Then if you look at the newer bug > https://debbugs.gnu.org/cgi/bugreport.cgi?bug=78473, that is on an > experimental branch which adds a new form of GC (not merged in master) > > So I think it's bad luck combined with it appearing in a struct that's > passed around often Right that might not be the same issue. There are other issues recently that has been exposed by SRA (PR 120987 for an example which was originally exposed by a SRA change but then reduced to remove the SRA part and found it was a latent bug earlier). Note GCC bug 119085 is definitely a dup and has a similar workaround as I have described in comment #19; the original testcase in comment #0 has the same issue with the union overlapping and using the wrong field for __total__ scalar replacement of aggregates when it came to unions.
