https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93776
--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Martin Jambor from comment #2) > But it is actually a slightly different problem. The assignment to > d7.pr, which is an empty struct, survives all the way to gimple as a > load from an initializer of size zero, which makes SRA to create an > access of size zero for it, which then confuses the verifier. Oh yes, the empty struct assignment issue from the gimplifier is recorded as PR 103645 which I have a fix for and will be submitting the patch soon.
