[Bug other/102440] Uinteger Opt/Param but the underlying type is signed

[segher at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102440

--- Comment #9 from Segher Boessenkool <segher at gcc dot gnu.org> ---
(In reply to Martin Liška from comment #8)
> > We could make the "UInteger" type mean it is implemented with an "unsigned
> > int"
> > C type (or some other unsigned integer type).
> 
> This would lead to the following list of -Wsign-compare warnings:
> 
> /home/marxin/Programming/gcc/gcc/c-family/c-opts.c:934:27: warning:
> comparison of integer expressions of different signedness: ‘unsigned int’
> and ‘int’ [-Wsign-compare]

That line is

  if (warn_shift_overflow == -1)

The documentation for that warning flag says
'-Wsign-compare'
     Warn when a comparison between signed and unsigned values could
     produce an incorrect result when the signed value is converted to
     unsigned.  In C++, this warning is also enabled by '-Wall'.  In C,
     it is also enabled by '-Wextra'.

I don't see how this could produce an incorrect result.  Writing the code as

  if (warn_shift_overflow == -1U)

means exactly the same thing, except it assumes the size of the variable so
it is a bad habit.  Plain "-1" is easier to read anyway.

It is idiom to use -1 for all-bits-set for unsigned vars.  It works correctly
whatever the size of the variable is.  It is silly if

  if (warn_shift_overflow == -1)

warns, but

  if (warn_shift_overflow + 1 == 0)

is just dandy (and that is the current situation :-( )

> One would need to verify/adjust all these places (plus many more for other
> targets).

Or fix the bloody warning ;-)  But not something you want on your plate, I
fully
understand :-)