instant team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Mon, Sep 01, 2008 at 02:00:02PM +0200, Martin Sandve Alnæs wrote:
> 2008/9/1 Anders Logg <logg@xxxxxxxxx>:
> > On Mon, Sep 01, 2008 at 01:20:29PM +0200, Martin Sandve Alnæs wrote:
> >> However, with two processes, there will still be a race
> >> condition during modification of the ~/.instant contents.
> >> Not sure if there's an easy way to handle that.
> >> We'll need some sort of locking mechanism on disk.
> >>
> >> Since two users will have different cache directories,
> >> this won't be a problem in that case, and better tempdir
> >> naming is enough.
> >
> > How likely is it that two people compile a form on the same machine
> > and make a call to Instant within the time it takes to create a
> > directory?
> >
> 
> Not very likely. But on e.g. marawa and potassium it may happen.
> If Instant is going to be as robust as possible, it should be handled.
> Not worth spending a lot of time on though...
> 
> However, the problem I state above can occur when a single user
> on a single computer runs two processes compiling the same form,
> where the form has not been cached before. One example comes to
> mind: Running Python MPI codes could lead to this problem.

ok.

-- 
Anders

Attachment: signature.asc
Description: Digital signature