On Thu, Jan 26, 2012 at 4:42 PM, Marc-André Lureau
<marcandre.lur...@gmail.com> wrote:
> On Thu, Jan 26, 2012 at 3:01 PM, Bruno Haible <br...@clisp.org> wrote:
>> Eric Blake wrote:
>>> <_timeval.h>:
>>> struct timeval
>>> {
>>> long tv_sec;
>>> long tv_usec;
>>> };
>>>
>>> Oops. struct timeval is generating a struct timeval with a 32-bit
>>> tv_sec even when time_t is 64-bits.
>>
>> mingw64 is still in development. Please, Marc-André, can you work with
>> them, upstream, to make sure that they follow the definition of
>> 'struct timeval' as described by POSIX [1][2]? This type is not defined
>> by the Windows API; the mingw people have the power to define it correctly.
Mingw64 reviewer said Windows actually defines it. I dig the msdn and
found this:
http://msdn.microsoft.com/en-us/library/windows/desktop/ms740560(v=vs.85).aspx
I suppose they are correct, and they should follow what Windows uses.
The problem was reported in an msdn discussion too:
http://social.msdn.microsoft.com/Forums/en/windowssdk/thread/674d34c9-b6f6-4380-bc7b-181eae99847a
Should gnulib redefine this struct similarly to other incompatible types?
--
Marc-André Lureau
