Re: lib/56275: pthread_cond_timedwait broken

[David Holland <dholland-bugs%netbsd.org@localhost>]

The following reply was made to PR lib/56275; it has been noted by GNATS.

From: David Holland <dholland-bugs%netbsd.org@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: lib/56275: pthread_cond_timedwait broken
Date: Fri, 25 Jun 2021 06:31:05 +0000

 On Fri, Jun 25, 2021 at 06:25:02AM +0000, David Holland wrote:
  > The following reply was made to PR lib/56275; it has been noted by GNATS.
  > 
  > From: David Holland <dholland-bugs%netbsd.org@localhost>
  > To: gnats-bugs%netbsd.org@localhost
  > Cc: 
  > Subject: Re: lib/56275: pthread_cond_timedwait broken
  > Date: Fri, 25 Jun 2021 06:21:38 +0000
  > 
  >  On Fri, Jun 25, 2021 at 01:30:01AM +0000, dholland%NetBSD.org@localhost wrote:
  >   >    then.tv_sec += ((unsigned long long)usecs * 1000 + start.tv_nsec) / 1000000000;
  >   >    then.tv_nsec = ((unsigned long long)usecs * 1000 + start.tv_nsec) % 1000000000;
  >  
  >  mlelstv points out: that += should be +
 
 or rather, what he actually said is that the += is wrong, and then
 because it's late I did the wrong thing and then wrote a different
 wrong thing here.
 
 	then.tv_sec = start.tv_sec;
 	then.tv_sec += ...
 	then.tv_nsec = ...
 
 and with that and the [thread %u trying cond %u] print commented out,
 on each run I get about a dozen spurious early wakeups after the
 initialization phase ends.
 
 -- 
 David A. Holland
 dholland%netbsd.org@localhost