Re: Testresults of libffi-3.0.9rc10 for x64 mingw target

[Andrew Haley <aph at redhat dot com>]

On 12/29/2009 08:01 AM, Kai Tietz wrote:
> 2009/12/28 Andrew Haley <aph@redhat.com>:
>> On 12/28/2009 07:15 PM, Kai Tietz wrote:
>>> 2009/12/28 Timothy Wall <twall@users.sf.net>:
>>>>
>>>> On Dec 28, 2009, at 12:50 PM, Andrew Haley wrote:
>>>>
>>>>> On 12/28/2009 05:16 PM, Anthony Green wrote:
>>>>>>
>>>>>> It looks like the failures are related to printing long long types.
>>>>>>
>>>>>> Andrew - do you remember the origin of this addition to ffitest.h...
>>>>>
>>>>> Yes.  See http://sourceware.org/ml/libffi-discuss/2009/msg00159.html
>>>>>
>>>>> Summary: mingw64 can't printf a long long.  According
>>>>> to Timothy Wall, "PRId64" is supposed to work on mingw64.
>>>>
>>>> I don't think "PRId64" is supposed to be in quotes, since it's a
>>>> macro, but I can't test atm.
>>>
>>> Indeed, PRId64 is a macro and has not to be quoted, as otherwise
>>> it simply gets a string, which isn't the intention here.
>>
>> This use of quotes looks perfectly alright to me.  What do you think
>> is supposed to be wrong with it?  The error messages suggest that
>> there is some problem interpreting the printf specifier.
>>
>> Index: testsuite/libffi.call/ffitest.h
>> ===================================================================
>> RCS file: /cvs/libffi/libffi/testsuite/libffi.call/ffitest.h,v
>> retrieving revision 1.5
>> diff -u -r1.5 ffitest.h
>> --- testsuite/libffi.call/ffitest.h     22 Dec 2008 15:21:15 -0000      1.5
>> +++ testsuite/libffi.call/ffitest.h     16 Jun 2009 09:37:58 -0000
>> @@ -45,6 +45,15 @@
>>
>>  #endif
>>
>> +/* MinGW kludge.  */
>> +#ifdef WIN64
>> +#define PRIdLL "PRId64"
>> +#define PRIuLL "PRIu64"
>> +#else
>> +#define PRIdLL "lld"
>> +#define PRIuLL "llu"
>> +#endif
> 
> Please take a look toe the PRIdLL for Non-WIN64 (and please use _WIN64
> as this is defined by compiler!!), and the Win64 version. Do you
> really believe that "PRId64" is a width specifier in a format string?

I assumed so.  The change was provided here:

http://sourceware.org/ml/libffi-discuss/2009/msg00175.html

Is this really the first time this has been tested?

> It isn't PRId64 is a macro defined in inttypes.h, which is a string,
> so please don't use quotes here in win64 case.

Sure, but we need someone to test it this time.

Andrew.