Bug #5933 for Time-HiRes: Time::HiRes::time does not pick up time adjustments like ntp

Mon Apr 05 12:06:19 2004 Guest - Ticket created

This 'bug' was found tracking down a problem I had with interval timers getting off track in the Event module. (This occurred under the Cygwin environment on a Windows 98 machine.) PCs are terrible time keepers. It seems that under heavy load, the clock can continually loose time. On my laptop, I run an NTP client every 5 minutes to compensate for this. Doing so then causes problems with Time::HiRes. Whereas Perl's core time() function picks up on the NTP clock adjustments, Time::HiRes::time() does not. To see this, run the following test script. #!/usr/bin/perl use strict; use warnings; use Time::HiRes; while (1) { print("Core : ", scalar localtime(), "\n"); print("HiRes: ", scalar localtime(Time::HiRes::time), "\n\n"); sleep(10); } Then adjust your PC's clock. You'll see that the 'Core' time picks up the adjustment, but the HiRes time does not.

Wed Apr 07 05:55:26 2004 JHI [...] cpan.org - Correspondence added

Yes, this is a side-effect of the implementation of Time::HiRes on Win32 - time adjustments made after the Perl script is started (or, rather, after the first time Time::HiRes::time() is called), are not reflected in the Time::HiRes::time(). Off-hand my Win32 informants do not promise any easy fixes since the high resolution of Time::HiRes::time() in Win32 is based on "performance counters" of Win32 which do not know anything about time adjustments like ntp.

Wed Apr 07 05:55:27 2004 JHI [...] cpan.org - Status changed from 'new' to 'open'

Wed Apr 07 05:56:39 2004 JHI [...] cpan.org - Subject changed from (no value) to 'Time::HiRes::time does not pick up time adjustments like ntp'

Mon Apr 19 14:30:17 2004 JHI [...] cpan.org - Correspondence added

I think Time::HiRes 1.59 should resolve the problem (as well as it is possible).

Mon Apr 19 14:30:18 2004 JHI [...] cpan.org - Status changed from 'open' to 'resolved'

Thu Apr 22 08:12:59 2004 Guest - Correspondence added

From:

jdhedden [...] 1979.usna.com

[JHI - Mon Apr 19 14:30:17 2004]: Show quoted text

> I think Time::HiRes 1.59 should resolve the problem (as well as it is > possible).

I tested this change. It appears that the drift rate may have decreased, but the problem still persists.

Thu Apr 22 08:13:00 2004 The RT System itself - Status changed from 'resolved' to 'open'

Sat May 08 04:05:59 2004 JHI [...] cpan.org - Correspondence added

This should have been fixed by 1.58 (1.59 is out by now), drift larger than 0.5 seconds should now be detected automatically by T::H and rectified.

Sat May 08 04:06:00 2004 JHI [...] cpan.org - Status changed from 'open' to 'resolved'

Sun May 09 10:55:35 2004 Guest - Correspondence added

From:

Jerry D. Hedden

[JHI - Sat May 8 04:05:59 2004]: Show quoted text

> This should have been fixed by 1.58 (1.59 is out by now), drift > larger than 0.5 seconds should now be detected automatically by > T::H and rectified.

It is not fixed. I have tested this twice in my system. In my most recent test, the clock actually started out 3 seconds behind. Over the course of 10 hours, it gained 20 seconds. My system uses Event.pm which, by default, uses Time::HiRes for its clock. When I configure Event.pm so that it uses core::Time, the clock tracks fine.

Sun May 09 10:55:36 2004 The RT System itself - Status changed from 'resolved' to 'open'

Wed May 12 16:19:46 2004 JHI [...] cpan.org - Correspondence added

Show quoted text

> It is not fixed. I have tested this twice in my system. In my most > recent test, the clock actually started out 3 seconds behind. Over > the course of 10 hours, it gained 20 seconds. My system uses Event.pm > which, by default, uses Time::HiRes for its clock. When I configure > Event.pm so that it uses core::Time, the clock tracks fine.

Rats. I am unfortunately out of ideas what to do since I do not have the Win32-fu to fix this.

Sat May 15 07:01:06 2004 Guest - Correspondence added

From:

Jerry D. Hedden

Here is a sample program and results that show the problem: #!/usr/local/bin/perl use strict; use warnings; use Time::HiRes; $| = 1; while (1) { sleep(900 - (CORE::time() % 900)); printf("Core: %02d:%02d:%02d HiRes: %02d:%02d:%02d\n", (localtime(CORE::time()))[2,1,0], (localtime(Time::HiRes::time()))[2,1,0]); } # EOF Results: Core: 22:00:00 HiRes: 22:00:00 Core: 22:15:01 HiRes: 22:15:00 Core: 22:30:13 HiRes: 22:29:59 Core: 22:45:00 HiRes: 22:44:46 Core: 23:00:01 HiRes: 22:59:46 Core: 23:15:01 HiRes: 23:14:45 Core: 23:30:01 HiRes: 23:29:44 Core: 23:45:00 HiRes: 23:44:43 Core: 00:00:01 HiRes: 23:59:43 Core: 00:15:00 HiRes: 00:14:42 Core: 00:30:00 HiRes: 00:29:42 Core: 00:45:00 HiRes: 00:44:42 Core: 01:00:01 HiRes: 00:59:42 Core: 01:15:01 HiRes: 01:14:41 Core: 01:30:02 HiRes: 01:29:40 Core: 01:45:00 HiRes: 01:44:38 Core: 02:00:02 HiRes: 01:59:38 Core: 02:15:01 HiRes: 02:14:36 Core: 02:30:01 HiRes: 02:29:35 Core: 02:45:01 HiRes: 02:44:34 Core: 03:00:02 HiRes: 02:59:33 Core: 03:15:01 HiRes: 03:14:31 Core: 03:30:02 HiRes: 03:29:30 Core: 03:45:01 HiRes: 03:44:28 Core: 04:00:01 HiRes: 03:59:27 Core: 04:15:02 HiRes: 04:14:26 Core: 04:30:01 HiRes: 04:29:24 Core: 04:45:02 HiRes: 04:44:23 Core: 05:00:01 HiRes: 04:59:21 Core: 05:15:02 HiRes: 05:14:20 Hope it helps.

Sun Dec 19 14:02:36 2004 JHI [...] cpan.org - Status changed from 'open' to 'resolved'