Paul Eggert wrote:
> > this platform's <time.h> contains a 'static inline'
> > definition of timespec_get
>
> That violates the C standard, as it means an extern inline function
> cannot call timespec_get.
Oh, indeed.
> Shouldn't we instead fix Gnulib's timespec_get module to work around
> this MSVC 14 bug? That way, all we need to do is have timespec_get
> callers depend on the timespec_get module, which they should do already.
The implementation of timespec_get for native Windows would be something
like
int
timespec_get (struct timespec *ts, int base)
{
/* Documentation:
https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/timespec-get-timespec32-get-timespec64-get1
*/
#ifdef _USE_32BIT_TIME_T
return _timespec32_get ((struct _timespec32 *) ts, base);
#else
return _timespec64_get ((struct _timespec64 *) ts, base);
#endif
}
But I am lazy here, as 'timespec_get' is usually not the kind of function
that people will call from an extern inline function.
Bruno
