Subject: Re: Errors with pkgsrc locking mechanism
To: Jaromir Dolecek <jdolecek@NetBSD.org>
From: Raymond Meyer <firstname.lastname@example.org>
Date: 07/30/2006 14:14:23
On Sun, 30 Jul 2006 13:36:41 +0200
Jaromir Dolecek <jdolecek@NetBSD.org> wrote:
> On Sun, Jul 30, 2006 at 12:15:10PM +0100, Raymond Meyer wrote:
> > On Sun, 30 Jul 2006 12:34:44 +0200
> > Lars Nordlund <email@example.com> wrote:
> > > I do not know what other people think about the pkgsrc locking
> > > mechanism, but IMHO it should be removed. Not only does it simply not
> > > work throughout pkgsrc because of the reasons stated above, I have also
> > > seen locks being grabbed when they should not. For example when the
> > > other make process has finished building and starts to install. Then the
> > > first make is released from the lock and starts to build.. With
> > > unspecified result of course... (from 2+ year old memory. Have not used
> > > it since then. Problem could have been corrected of course).
> > >
> > >
> > > Best regards,
> > > Lars Nordlund
> > I didn't know that pkgsrc locking was unsafe. What's the point of having
> > something if it only works properly 80% of the time, or so.
> I'm a PKGSRC_LOCKTYPE=sleep user here, and it works fine for me -
> I've not seen any lock-related problems so far.
Then you must have not seen the mail I sent to this list that started this
thread. I was building packages on a 4-way Sun E450 machine and I encountered
problems with pkgsrc locking. Do you use pkgsrc on SMP hardware? How often do
you build packages in parallel?