https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128638

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
And for value-numbering this shows that the loads in

  if (x)
    memcpy(&t, p, 4);
  else
    t = *(int *)p;

both get associated to the same memory state from the end of the conditional
block.  That reverting r12-3918 fixes the testcase is spurious and solely
depends on the visitation order of the two blocks.

Reply via email to