ffc team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Fri, Feb 27, 2009 at 10:03:06AM +0100, Anders Logg wrote:
> On Fri, Feb 27, 2009 at 09:45:49AM +0100, Kristian Oelgaard wrote:
> > Quoting buildbot@xxxxxxxxxx:
> > 
> > > The Buildbot has detected a new failure of ffc-hardy-i386 on FEniCS
> > > Buildbot.
> > > Full details are available at:
> > >  http://fenics.org:8080/builders/ffc-hardy-i386/builds/42
> > > 
> > > Buildbot URL: http://fenics.org:8080/
> > > 
> > > Buildslave for this Build: hardy-i386
> > > 
> > > Build Reason: 
> > > Build Source Stamp: HEAD
> > > Blamelist: 
> > > 
> > > BUILD FAILED: failed ffc check
> > 
> > The JIT unit test fails as I reported earlier when I do:
> > 
> > instant-clean
> > python test.py
> > 
> > but if I run
> > python test.py
> > 
> > a second time the test is OK.
> > Kristian

The problem is that the first time a form is compiled, FFC will
simplify the form and thus the signature will be different on the
second call. I solved this by jit-compiling twice before timing.

Don't know why it worked before when instant-clean has not been
called. It should only affect the disk cache. There is probably a good
explanation but I haven't bothered to figure it out. It will change
with UFL anyway since the compiler will not be allowed to modify the
forms.

-- 
Anders

Attachment: signature.asc
Description: Digital signature