On Wed, Jul 01, 2026 at 09:47:46AM +0200, Eric Botcazou wrote: > this is a regression present on mainline, 16, 15 and 14 branches introduced > by > the fix for PR tree-optimization/112653 (PTA and return). What happens is > that DSE incorrectly eliminates a call to __builtin_memcpy, whose destination > is obtained from (an equivalent of) malloc and is ultimately returned from > the > function. But this happens only when the dynamic allocation is conditional.
For us Ada illiterate, here is a C++ testcase which got fixed by this too. Ok for trunk and perhaps release branches too? 2026-07-03 Jakub Jelinek <[email protected]> * g++.dg/opt/20260703-1.C: New test. --- gcc/testsuite/g++.dg/opt/20260703-1.C.jj 2026-07-03 17:01:03.845156091 +0200 +++ gcc/testsuite/g++.dg/opt/20260703-1.C 2026-07-03 17:00:48.744347853 +0200 @@ -0,0 +1,46 @@ +// This started to be miscompiled with r15-579 or in a larger +// test with r15-3956 and got fixed with r17-2039. +// DSE would optimize away the store of 42. +// { dg-do run { target c++11 } } +// { dg-options "-O2" } + +struct B { }; + +void *volatile g; + +template <typename T> +struct D { T foo (void) const { return (T) g; } }; + +struct G : public B +{ + virtual B *bar (int); + struct H : public B { H () : h (42) {} int h; }; + struct I : public B {}; + D <I *> g; +}; + +struct H : public G { virtual B *bar (int) { return nullptr; } }; + +B * +G::bar (int x) +{ + if (x == 0) + return new H (); + I *y = g.foo (); + return y; +} + +[[gnu::noipa]] B * +baz (G *x, int y) +{ + return x->bar (y); +} + +int +main () +{ + G g; + G::H *h = (G::H *) baz (&g, 0); + if (h->h != 42) + __builtin_abort (); +} Jakub
