Re: src/winsup/utils ChangeLog.64bit cygcheck.cc l ...

[Eric Blake <eblake at redhat dot com>]

On 02/06/2013 10:57 AM, Corinna Vinschen wrote:

>> -      printf ("ldd: GetModuleFileName returned an error %lu\n",
>> GetLastError ());
>> +      printf ("ldd: GetModuleFileName returned an error %" PRIu32 "\n",
>> +	      (unsigned int) GetLastError ());
>>
>> PRIu32 is for uint32_t, not unsigned int; and on platforms where
>> uint32_t is unsigned long, you've ended up causing a compiler warning.
>> If you are going to cast GetLastError(), then just use %u; no need to
>> drag in <inttypes.h>.
> 
> I changed that in ldd.cc.  Other than that, We *know* that DWORD is
> declared as unsigned long on 32 bit, but as unsigned int on 64 bit,
> because sizeof long == 64 on x86_64-pc-cygwin.  Therefore I'm using
> the PRIu32 and PRIx32 to make sure that I have the right format
> specifier for a known 32 bit variable, which uses a different type
> on different platforms.

If DWORD were defined as uint32_t, then this would work without
questions.  But because DWORD is defined as 'unsigned long' vs.
'unsigned int', rather than uint32_t, your trick of using PRIu32 will
only work if we also ensure that uint32_t is defined to the same type as
DWORD.  I guess we could get away with that, but it is rather subtle.
If we go with that approach, then it would be sufficient to write:

 printf("%" PRIu32 "\n", GetLastError ())

But if we are worried that DWORD and uint32_t can ever be defined
independently, then we risk mismatch between long and int on 32-bit
platforms, even though we know both types are 32-bits, and it is that
type mismatch that gcc warns about.

Anyways, thanks for the followup patch.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

Attachment: signature.asc
Description: OpenPGP digital signature