Re: MPQS not cleaning up files

[Bill Allombert <allomber@math.u-bordeaux.fr>]

On Thu, Oct 03, 2002 at 01:09:05PM -0400, Igor Schein wrote:
> On Thu, Oct 03, 2002 at 06:39:40PM +0200, Karim BELABAS wrote:
> > On Wed, 2 Oct 2002, Michael Somos wrote:
> > > -rw-------    1 somos    somos        7224 Jul 21 15:04 FNEW.614.15591
> > > -rw-------    1 somos    somos      220962 Jul 21 15:04 FREL.614.15591
> > > -rw-------    1 somos    somos        8055 Jul 21 15:04 LPNEW.614.15591
> > > -rw-------    1 somos    somos      192510 Jul 21 15:04 LPREL.614.15591
> [snip]
> > (2) restart interrupted jobs in case of system shutdown: it is not uncommon to
> > let MPQS run for a few months.
> > 
> > It has been a long time since a bug was reported in this module
> > [ unfortunately, development stopped at about the same time ]. So (1) is
> > probably no more a concern. (2) still is, although with the current state of
> > the interface, it would be difficult to recover an interrupted job (this is
> > not documented in any case).
> 
> I've done a restart before - very easy.  I should still have Gerhard's
> instructions.  I'd rather leave the behavior unchanged for such cases
> where it has run for too long to be wasted.

I should add two things:

1) If you do a Ctrl-C to stop a factorisation, it is dubious that you will want to
restart it.

2) This is currently in the /tmp directory, which can be cleaned on reboot, so
this offers no protections against system crash.

Also accessing /tmp is sometime faster than /var/tmp.  Probably the most clever
solution would be to use /tmp for short factorisation (say <60 digits) and
/var/tmp for larger factorization. Unfortunately for a software being 'clever'
is sometime close to be 'braindead'.

Maybe we need a clear Configure option or even a default.
And document the recovery procedure.

Cheers,
Bill.