Re: [Qemu-devel] [5275] Suppress gcc 4.x -Wpointer-sign (included in -Wall) warnings

[Gerd Hoffmann <kraxel@redhat.com>]

Andreas Färber wrote:
> 
> Am 28.10.2008 um 17:54 schrieb Ian Jackson:
> 
>> Blue Swirl writes ("[Qemu-devel] [5275] Suppress gcc 4.x
>> -Wpointer-sign (included in -Wall) warnings"):
>>> -        ret = sprintf(phys_ram_base + (16 << 20) - 256,
>>> +        ret = sprintf((char *)(phys_ram_base + (16 << 20) - 256),
>>
>> I realise I'm a bit late with this comment (my apologies), but:
>>
>> I think this shows that it would be better to suppress the warning
>> with the appropriate compiler option, than to try to update the code.

/me agrees.

>> And these kind of casts are dangerous because they can suppress
>> serious warnings about pointer/integer mismatch.

Exactly that's why.

> Not sure what type phys_ram_base had here, but char can be signed or
> unsigned depending on platform and this often leads to warnings on,
> e.g., OSX.

Pointer sign-ess for char is a PITA *because* just "char" can be signed
or unsigned depending on the platform.  IIRC that is the reason the
linux kernel has the warning turned off.

> Suppressing all signedness warnings could hide real mistakes,
> too...

That is true too.  Unfortunaly gcc has no -Wno-pointer-sign-char

cheers,
  Gerd