Dustin Sallings wrote:
> On Sep 27, 2010, at 5:43, Simon Slavin wrote:
>
>> I think this differs for different platforms. For instance:
>>
>> http://www.manpagez.com/man/3/usleep/
>>
>> makes no mention of that fault.
>
> It kind of does indirectly. It says it's built on nanosleep(), and
> links to that doc which says this in their error section:
>
> [EINVAL] rqtp specified a nanosecond value less than zero or
> greater than or equal to 1000 million.
>
> I didn't actually run into a problem with it, I just noticed that its
> result isn't checked and documentation I was reading had it defined to fail.
>
>> Since the time for sleeping is an integer number of seconds it's probably
>> simplest just to replace this call with one to sleep(10). sleep() is
>> defined in POSIX.1 so I believe it should be available in a library for
>> every platform SQLite is expected to run on.
>
> I think that makes sense. It also means you don't need a
> comment describing that it's intended to sleep for exactly ten seconds. :)
Yes, that might be a problem.
I see than in Vim source code, usleep() is used as follows
to avoid calling usleep() with a large argument:
#ifdef HAVE_NANOSLEEP
...
#else
# ifdef HAVE_USLEEP
while (msec >= 1000)
{
usleep((unsigned int)(999 * 1000));
msec -= 999;
}
usleep((unsigned int)(msec * 1000));
# else
...
#endif
-- Dominique
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
