[Bug tree-optimization/125419] [16/17 Regression] ICE at -O3 during GIMPLE pass: ifcvt

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125419

--- Comment #5 from Drea Pinski <pinskia at gcc dot gnu.org> ---
Comment on attachment 64529
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64529
Patch which I am testing

>-  /* At this point there should be no ssa names occuring in abnormals.  */
>-  gcc_assert (!arg0_op.operands_occurs_in_abnormal_phi ());
>+  /* Might pick up abnormals from previous bbs so stop the loop.  */
>+  if (!arg0_op.operands_occurs_in_abnormal_phi ())
>+    return;

This should have been `if (arg0_op.operands_occurs_in_abnormal_phi ())`

> 
>   gimple *arg1_def_stmt = SSA_NAME_DEF_STMT (*arg1);
>   if (!gimple_extract_op (arg1_def_stmt, &arg1_op))
>     return;
> 
>-  /* At this point there should be no ssa names occuring in abnormals.  */
>-  gcc_assert (!arg1_op.operands_occurs_in_abnormal_phi ());
>+  /* Might pick up abnormals from previous bbs so stop the loop.  */
>+  if (!arg1_op.operands_occurs_in_abnormal_phi ())
>+    return;

Likewise.

Testing with that corrected rather than the original patch; Noticed just after
testing.