https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109596
--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #9)
> BTW, just curious, r14-162 had:
> /* Ensure that the header will have just the latch as a predecessor
> inside the loop. */
> - if (!single_pred_p (exit->dest))
> + if (!single_pred_p (nonexit->dest))
> {
> - header = split_edge (exit);
> + header = split_edge (nonexit);
> exit = single_pred_edge (header);
> }
> chunk, shouldn't that be nonexit = single_pred_edge (header); or is that
> line correct?
Regarding that, e.g. gcc.c-torture/compile/pr70199.c testcase at -O1 -g shows
both cases, once where single_pred_p (nonexit->dest) and once where
!single_pred_p (nonexit->dest) and we split the nonexit edge.
Or the !single_pred_p case is also in c-c++-common/gomp/pr59917-2.c,
gcc.dg/torture/pr83685.c.