Re: Idea to refine -Wtype-limits

[Gabriel Dos Reis <gdr at cs dot tamu dot edu>]

Gerald Pfeifer <gerald@pfeifer.com> writes:

| -Wtype-limits guards a very useful, in my experience, set of warnings
| including the following:
| 
|   warning: comparison of unsigned expression < 0 is always false
|   warning: comparison of unsigned expression >= 0 is always true
| 
| Based on these I identified and fixed several real bugs in Wine so far
| and I'd like to enable -Wtype-limits by default.  However, there is one 
| caveat:  -Wtype-limits also warns about seemingly harmless code like
| 
|   bool f(unsigned i) {
|     if( DEFINE_FIRST <= i || i <= DEFINE_LAST )
|       return false;
|     return true;
|   }
| 
| in case DEFINE_FIRST evaluates to 0.
| 
| In applications like Wine there are a couple of such cases and it would 
| not be good to remove the first half of the check to silence the warning 
| because at a later point in time the DEFINE_FIRST might actually change,
| plus the simplification is not obvious looking at the source.
| 
| So, I got a "creative" idea.  What if we disable this warning if comparing 
| against 0U instead of 0?  Then we could use #define DEFINE_FIRST 0U and 
| avoid the warning for cases like this, without losing real diagnostics
| assuming that nobody would accidently write such code.
| 
| Ah, and we wouldn't even have to adjust any of the messages in GCC, just 
| the check itself. ;-)

:-)

| Thoughts?

I have no implementation strategy at the moment.  But I do have a
request:  Please, whatever you do, please, don't make it noisy with 
template codes and `reasonable' codes. (We have had PRs for that
behaviour in template codes).

-- Gaby