https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108868
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
Status|NEW |ASSIGNED
--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
Btw, there's also cleanup_call_ctrl_altering_flag in CFG cleanup which cleans
up the flag on the _last_ stmt of blocks. Dependent on whether fork/setjmp
is last it would clean the flag based on outgoing edges not being abnormal.
That's another reason to not abuse the ctrl-altering flag here.
I'm going to do the reversion now - is there a runtime testcase we can add for
this PR or should we solely rely on the abnormal dispatcher presence?
Btw, if the testcase would be just
void foo ()
{
if (!vfork ())
_exit (0);
}
then it's IMHO OK to elide the abnormal edges? There's still the issue
of clobbering if RTL doesn't have a REG_SETJMP note on the vfork (it
lacks the abnormal edges preventing code motion I think).
I'm going to revert the offending rev. now.