https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101941
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Target Milestone|--- |12.0
Status|UNCONFIRMED |NEW
CC| |hubicka at gcc dot gnu.org
Ever confirmed|0 |1
Last reconfirmed| |2021-08-17
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
Yes, this is PR94818. There's always going to be cases like this - it could be
RTL optimizations eliding the last call. We're emitting functions in postorder
so optimizations there improve analysis on the call side. There's no perfect
order and since the error in question is emitted by RTL expansion it's too
early anyway.
Now, we could heuristically avoid splitting away error paths which would
mitigate the situation a bit I guess. Honza?
