[Bug c/104205] New: Signed overflow in case label reports confusing warning with -Wpedantic

[Bug c/104205] New: Signed overflow in case label reports confusing warning with -Wpedantic

[stephen.wassell at sophos dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104205

            Bug ID: 104205
           Summary: Signed overflow in case label reports confusing
                    warning with -Wpedantic
           Product: gcc
           Version: 11.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: stephen.wassell at sophos dot com
  Target Milestone: ---

When compiling the following code with "gcc -Wpedantic":

    int convert(int err)
    {
        switch (err)
        {
            case 1 << 31: // signed overflow - undefined behaviour
                return 0;
            default:
                return 1;
        }
    }

    int main()
    {
        return convert(1 << 31);
    }

It reports a confusingly worded warning since version 6. It should say
something about undefined behaviour instead. Should there also be a warning for
the other use of 1 << 31?

    <source>: In function 'convert':
    <source>:5:9: warning: case label is not an integer constant expression
[-Wpedantic]
    5 |         case 1 << 31:
      |         ^~~~

Seen on gcc (Ubuntu 11.2.0-7ubuntu2) 11.2.0 x64, and multiple versions on
Compiler Explorer.

https://godbolt.org/z/4Tr1qxaqv

[Bug c/104205] Signed overflow in case label reports confusing warning with -Wpedantic

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104205

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|https://godbolt.org/z/4Tr1q |
                   |xaqv                        |
           Keywords|                            |diagnostic

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
https://godbolt.org/z/4Tr1qxaqv

[Bug c/104205] Signed overflow in case label reports confusing warning with -Wpedantic

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104205

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
>  It should say something about undefined behaviour instead.

The problem is for integer constants, it is not undefined at runtime but rather
invalid code at compile time.

>Should there also be a warning for the other use of 1 << 31?

I think you misunderstood the warning/error fully.

In C99-C17, 1<<31 is not an integer constant expression as there is an
overflow. 

But it is still valid, just undefined at runtime if used where an integer
constant expression is not required. case statements require an integer
constant expression.

[Bug c/104205] Signed overflow in case label reports confusing warning with -Wpedantic

[stephen.wassell at sophos dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104205

--- Comment #3 from Stephen Wassell <stephen.wassell at sophos dot com> ---
Thank you for the response! I think the cause of our confusion was that we were
focusing on "constant" in the warning message rather than "integer". Our
original code had a few layers of macros so the signed overflow wasn't obvious
at first.

[Bug c/104205] Signed overflow in case label reports confusing warning with -Wpedantic

[egallager at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104205

Eric Gallager <egallager at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |egallager at gcc dot gnu.org

--- Comment #4 from Eric Gallager <egallager at gcc dot gnu.org> ---
Maybe the warning text could be updated to include what's actually being used
for the case label after macro expansion etc.? e.g.:
<source>:5:9: warning: case label '1 << 31' is not an integer constant
expression [-Wpedantic]

[Bug c/104205] Signed overflow in case label reports confusing warning with -Wpedantic

[joseph at codesourcery dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104205

--- Comment #5 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
Printing an expression within the diagnostic message is generally 
problematic, but it might be good to change the caret location to point to 
the expression rather than the keyword "case" (or to a range covering both 
the keyword and the expression).