https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119903
Bug ID: 119903
Summary: [13/14/15/16 Regression] non-call exception vs traping
comparison
Product: gcc
Version: 16.0
Status: UNCONFIRMED
Keywords: EH, wrong-code
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: pinskia at gcc dot gnu.org
Reporter: pinskia at gcc dot gnu.org
Target Milestone: ---
Take:
```
int f() noexcept;
int g() noexcept;
int m(double a)
{
try {
if (a < 1.0)
return f();
return g();
}catch(...)
{
return -1;
}
}
```
Compile with `-O2 -fnon-call-exceptions -ftrapping-math`, we lose the exception
landing pad in forwprop1.
And we get:
```
;; basic block 2, loop depth 0, count 1073741824 (estimated locally), maybe
hot
;; prev block 0, next block 3, flags: (NEW, VISITED)
;; pred: ENTRY [always] count:1073741824 (estimated locally)
(FALLTHRU,EXECUTABLE)
[LP 1] _5 = a_4(D) < 1.0e+0;
;; succ: 7 [never] count:0 (precise) (EH,EXECUTABLE)
;; 3 [always] count:1073741824 (estimated locally)
(FALLTHRU,EXECUTABLE)
;; basic block 3, loop depth 0, count 1073741824 (estimated locally), maybe
hot
;; prev block 2, next block 4, flags: (NEW, VISITED)
;; pred: 2 [always] count:1073741824 (estimated locally)
(FALLTHRU,EXECUTABLE)
_15 = a_4(D) < 1.0e+0;
if (_15 != 0)
```
Notice that _5 is no used so it will be DCE (on the RTL level).
This started with r13-322-g7f04b0d786e13f.
I have a "hack" to gimple-fold to fix this by rejecting what match and simplify
has done after the fact. But I almost think we should reject this kind of thing
outside of gimple fold too.
