[Bug c++/94767] New: (unsigned bitfield) + (int) operation results in int, not unsigned int.

[Bug c++/94767] New: (unsigned bitfield) + (int) operation results in int, not unsigned int.

[jh718.park at samsung dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94767

            Bug ID: 94767
           Summary: (unsigned bitfield) + (int) operation results in int,
                    not unsigned int.
           Product: gcc
           Version: 9.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: jh718.park at samsung dot com
  Target Milestone: ---

For these variables below,

unsigned m_schemeEnd : 26;
unsigned m_userStart;

m_userStart == m_schemeEnd + 1

this comparison emits a compiler warning as below.

warning: comparison of integer expressions of different signedness: ‘unsigned
int’ and ‘int’ [-Wsign-compare]

This bug was found during WebKit gtk port build with gcc 9.3.0.

Temporarily, this warning was removed by this patch,
https://trac.webkit.org/changeset/260715/webkit
like

bool slashSlashNeeded = m_userStart == static_cast<unsigned>(m_schemeEnd + 1);

but we think that this bug should be fixed in gcc.

[Bug c++/94767] (unsigned bitfield) + (int) operation results in int, not unsigned int.

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94767

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I think this has been fixed on the trunk.

[Bug c++/94767] (unsigned bitfield) + (int) operation results in int, not unsigned int.

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94767

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |diagnostic, wrong-code
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2020-04-27

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Who knows, please provide a testcase that reproduces the issue.

[Bug c++/94767] (unsigned bitfield) + (int) operation results in int, not unsigned int.

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94767

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to jh718.park from comment #0)
> For these variables below,
> 
> unsigned m_schemeEnd : 26;
> unsigned m_userStart;
> 
> m_userStart == m_schemeEnd + 1
> 
> this comparison emits a compiler warning as below.
> 
> warning: comparison of integer expressions of different signedness:
> ‘unsigned int’ and ‘int’ [-Wsign-compare]

Why do you think that's wrong?

In [conv.prom] p5 the standard says:

"A prvalue for an integral bit-field (11.4.9) can be converted to a prvalue of
type int if int can represent all the values of the bit-field;"

Since int is wider than 26 bits it can represent all the values of m_schemeEnd,
so the operands of 'm_schemeEnd + 1' are both promoted to int, and the result
is an int.

[Bug c++/94767] (unsigned bitfield) + (int) operation results in int, not unsigned int.

[jh718.park at samsung dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94767

jh718.park at samsung dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|---                         |INVALID

--- Comment #4 from jh718.park at samsung dot com ---
(In reply to Jonathan Wakely from comment #3)
> (In reply to jh718.park from comment #0)
> > For these variables below,
> > 
> > unsigned m_schemeEnd : 26;
> > unsigned m_userStart;
> > 
> > m_userStart == m_schemeEnd + 1
> > 
> > this comparison emits a compiler warning as below.
> > 
> > warning: comparison of integer expressions of different signedness:
> > ‘unsigned int’ and ‘int’ [-Wsign-compare]
> 
> Why do you think that's wrong?
> 
> In [conv.prom] p5 the standard says:
> 
> "A prvalue for an integral bit-field (11.4.9) can be converted to a prvalue
> of type int if int can represent all the values of the bit-field;"
> 
> Since int is wider than 26 bits it can represent all the values of
> m_schemeEnd, so the operands of 'm_schemeEnd + 1' are both promoted to int,
> and the result is an int.

Thank you for your comment, Jonathan.
I understand your point.

I thought that unsigned bitfield should be converted to unsigned
during usual arithmetic conversions without knowing the item,
http://eel.is/c++draft/conv.prom#5.

Then, I will mark this issue as resolved/invalid, and update the bug
https://bugs.webkit.org/show_bug.cgi?id=211044
with the comment you left here.

Thank you for your help:)