Re: [PATCH] cooker: release lockfile on process exit

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Tue, 2015-05-19 at 18:24 -0300, Otavio Salvador wrote:
> On Tue, May 19, 2015 at 6:07 PM, Richard Purdie
> <richard.purdie@linuxfoundation.org> wrote:
> ...
> > So having thought more about this, my proposal is actually that rather
> > than fix this and cause all kinds of other potential problems, we just
> > make memory resident bitbake the default which is something we'd like to
> > ideally do in the 1.9 timeframe anyway.
> 
> I understand your background ideas here however the non-deterministic
> lock release is driving us crazy at the autobuilder management. The
> sstate cleanup script, workdir cleanup script and in-house scripts all
> need to be teach to deal with this and this is far from optimal.
> 
> Would be acceptable for you a command line option to 'wait' it to
> release the lock before exit? This could be done easily and we avoid
> side effects. This also provide a solution which is isolated enough to
> be safely backported for previous releases and provide more
> deterministic script output.
> 
> How does it sounds?

I don't know how you'd implement this safely. I did give this some
further thought last night and concluded that if we want to backport
something, the safest approach might be to change the lock acquisition
code.

Basically, if we can't obtain the lock, sleep for one second and try
again. If on the second try we can't obtain the lock, error/exit.

If we need a longer delay than one second, we need to figure out what is
exiting that slowly and fix that instead.

Seems reasonable?

Cheers,

Richard