[Bug tree-optimization/119930] [16 regression] g++.dg/coroutines/torture/pr103953.C FAILs with -O3

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119930

--- Comment #20 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #19)
> (In reply to Andrew Pinski from comment #18)
> > (In reply to Richard Biener from comment #17)
> > > I'm not sure why this was marked as 15 regression, 15.1 works for me, so
> > > does 15.2.
> > 
> > Because the slightly modified testcase in comment #8 is failing in GCC
> > 15.1.0 too. the only difference between this testcase and the testcase in
> > the testsuite is the main is renamed to f and then main calls f.
> 
> Meaning r16-89 exposed the latent bug in GCC 15 and bisecting the testcase
> in comment #8 gets us to r15-5595 . I have not figured out why
> -fassume-sane-operators-new-delete would cause the failure yet.

Conjecture - for a coroutine frame it would be an invalid analysis to assume
that a paired new and delete could be removed [unless/until we implement
heap-allocation elision].  Such analysis should be disabled for coroutine
functions.