[Bug tree-optimization/104356] [12 Regression] divide by zero trap is being removed now when it should not be in some cases

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104356

--- Comment #32 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Comment on attachment 52340
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=52340
Tentative fix

For the X / bool_range_Y is X. case I think just !flag_non_call_exceptions
would be better.  If @1 has boolean range and is known to be non-zero, it is
known to be 1, so we should be optimizing it elsewhere, that is the constant
case.

Anyway, that's probably it in match.pd.
path isolation is already guarding is_divmod_with_given_divisor calls with
!cfun->can_throw_non_call_exceptions.
>From the to me known issues that leaves VRP.  As Ada expects the division by
zero to trap, I guess we don't care much about the range it emits for the
result, but it shouldn't be for cfun->can_throw_non_call_exceptions be
undefined or single value range, because then it would be just optimized into a
constant I bet.  But there is also the problem that ranger doesn't walk just
from stmts to its immediate uses, but also from stmts to their operand's def
stmts.
Can you try to rewrite the
unsigned                                                                        
foo (unsigned x, unsigned y)                                                    
{                                                                               
  if (x >= 2)                                                                   
    return 0;                                                                   
  if (x == 1)                                                                   
    y += 4;                                                                     
  return y / x;                                                                 
}                                                                               
testcase I've posted into Ada and see if it will optimize away the division in
evrp or vrp?