[Bug tree-optimization/126856] [17 Regression] mpfr and mpc started failing tests recently since r17-3204

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126856

Drea Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
            Summary|[17 Regression] mpfr and    |[17 Regression] mpfr and
                   |mpc started failing tests   |mpc started failing tests
                   |recently                    |recently since r17-3204
                 CC|                            |pinskia at gcc dot gnu.org
   Last reconfirmed|                            |2026-08-13

--- Comment #1 from Drea Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

What is happening is we have:
  exp.0_1 = (unsigned int) exp_8(D);

Which gets recorded as:
Folding statement: exp.0_1 = (unsigned int) exp_8(D);
 Registering value_relation (exp.0_1 pe32 exp_8(D)) (bb2) at exp.0_1 =
(unsigned int) exp_8(D);


and then when the simple_dce happens we change the assignment to exp.0_1 to:

exp.0_1 = 0;

And then we look at:
Folding statement: _5 = (short unsigned int) exp_8(D);
 Registering value_relation (_5 pe16 exp_8(D)) (bb6) at _5 = (short unsigned
int) exp_8(D);
Queued stmt for removal.  Folds to: 0

Which is wrong.

So this is caused by r17-3204-ge02ce3b8c4574c.

Let me try to get a testcase without unions and fp.