This is the mail archive of the
mailing list for the Cygwin project.
Re: pthread_cond_timedwait accurate to one second only
- From: "Timothy C Prince" <tprince at myrealbox dot com>
- To: pgarrone at linuxmail dot org
- Cc: cygwin at cygwin dot com
- Date: Mon, 25 Aug 2003 15:26:14 +0000
- Subject: Re: pthread_cond_timedwait accurate to one second only
From: "peter garrone" <email@example.com>
Date: Mon, 25 Aug 2003 15:42:52 +0800
Subject: pthread_cond_timedwait accurate to one second only
I would like to use this function down to 10 milliseconds accuracy if possible.
However upon looking at winsup/cygwin/thread.cc,
it uses the function "ftime" and the millisecond field is ignored.
All the examples in the winsup testsuite also generally check to 5 seconds only.
Is there any inherent reason why finer timing would not work?
In order to avoid going to newlib and implementing Windows GetLocalTime() calls in gettimeofday(), we made the -mwin32 build option invoke the API milliseconds call directly in the g77 runtime libf2c/libU77/datetime_.c. I think the "inherent reason" is the perceived awkwardness in making <sys/time.h> library calls pick up milliseconds on Windows, while leaving the lower order part of the microseconds field undetermined. My documentation on ftime() says it is obsoleted by gettimeofday().
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html