[Bug middle-end/104787] [12/13 Regression] False positive -Wreturn-type since r12-5638-ga3e75c1491cd2d50

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104787

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2
                 CC|                            |rguenth at gcc dot gnu.org

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
So the case is probably similar - without the early folding we go all the
way, transforming

    struct FileAccessRef f;
  try
    {
      if (1)
        {
          return <retval> = 0;
        }
    }
  finally
    {
      FileAccessRef::~FileAccessRef (&f);
    }
  __builtin_trap ();

to

D.2402 = 0;
// predicted unlikely by early return (on trees) predictor.
finally_tmp.0 = 0;
FileAccessRef::~FileAccessRef (&f);
if (finally_tmp.0 == 1)
  goto <bb 4>; [INV]
else
  goto <bb 6>; [INV]

since the return-type diagnostic is before any kind of optimization it
probably diagnoses this by design.  Otherwise it would be appropriate
to move the diagnostic to the build_ssa_passes set of diagnostic passes
which could share a common VN run to mark unreachable blocks.

In the end I'm currently not working on the middle-end unreachable code
diagnostic so we can as well revert the frontend change causing all this
fallout ...