[Bug c++/102303] New: optimizer elides -ftrapv overflow detection

[Bug c++/102303] New: optimizer elides -ftrapv overflow detection

[john at mcfarlane dot name]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102303

            Bug ID: 102303
           Summary: optimizer elides -ftrapv overflow detection
           Product: gcc
           Version: 11.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: john at mcfarlane dot name
  Target Milestone: ---

With flag, `-ftrapv`, the following program on CE with 'x86-64 gcc 11.2' has
exit status 139:

   int main()
   {
       int n = 0x7fffffff;
       return n+1;
   }

This is expected behavior: run-time overflow trapping. But with flags, `-ftrapv
-O1`, it has exit status 0. I wouldn't expect optimization to change the
behavior of this program, which appears to be either UB or as-if `-fwrapv`.

https://godbolt.org/z/jvj9javcY

Behavior is inconsistent with Clang 12.0. Behavior is inconsistent with
`-fsanitize=signed-integer-overflow -fsanitize-undefined-trap-on-error`.

Inconsistency goes as far back as the 5.x series, where trapping of this
overflow first appears. Docs for -ftrapv
(https://gcc.gnu.org/onlinedocs/gcc/Code-Gen-Options.html) don't mention
optimization.

[Bug c++/102303] optimizer elides -ftrapv overflow detection

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102303

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This is a known bug with -ftrapv where gcc widens everything to word mode
(64bit in this case) and uses libcalls to implement-ftrapv. There are a few
other bugs that say the same thing.

[Bug c++/102303] optimizer elides -ftrapv overflow detection

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102303

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
No, the issue is that we apply constant folding and elide the trapping
operation which is generally considered OK by GCC (eliding traps, that is). 
For example
we also elide the trap in

   int main()
   {
       int n = 0x7fffffff;
       n+1;
       return 0;
   }

I would advise against the use of -ftrapv and instead use
-fsanitize=signed-integer-overflow -fsanitize-undefined-trap-on-error

[Bug c++/102303] optimizer elides -ftrapv overflow detection

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102303

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |DUPLICATE
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This is a dup of bug 61893.

*** This bug has been marked as a duplicate of bug 61893 ***