bug#22444: compiler warnings in test snapshot

[Charles Diza]

On Sat, Jan 23, 2016 at 9:10 PM, Paul Eggert <address@hidden> wrote:

> Thanks for checking this so quickly!
>
> Charles Diza wrote:
>
> error.c:386:12: warning: data argument not used by format string
>>
>
> That one is a false alarm. The format string might contain fewer formats
> than there are arguments, and C allows this so the behavior is
> well-defined. Writing it differently would hurt performance slightly.
>
>
> CC       strstr.o
>> In file included from regex.c:70:
>> ./regex_internal.c:1392:11: warning: comparison of unsigned expression < 0
>> is
>>        always false [-Wtautological-compare]
>>    if (idx < 0 || idx >= set->nelem)
>>        ~~~ ^ ~
>>
>
> Another false alarm. On some configurations idx is signed and so the test
> is needed there. It's harmless (no code generated) on your platform, where
> idx is unsigned.
>
> I assume these are Clang warnings.
>

Yes, these are clang warnings.


>
> Does the second warning go away if the code looks like this?
>
>   if (((Idx) -1 < 0 && idx < 0) || idx >= set->nelem)
>

No, the second warning is still there.


> Dunno if it's worth this sort of obfuscation just to pacify Clang, tho....
>

Probably not.  Now that you have confirmed that the warnings are harmless,
they don't bother me.

Cheers,
Charles