I just browsed through the changesets committed since the release and searched for "parameters" and "form_compiler_parameters" and found several hits here:
https://bitbucket.org/fenics-project/dolfin/commits/74b0f05ebdd16003749674482e2eb46f4dadc7cc?at=next I'm not saying I see anything wrong, just that it's code that could possibly result in a KeyError: "'form_compiler'". -- Anders On Wed, Jun 04, 2014 at 10:22:35AM +0200, Martin Sandve Alnæs wrote: > How do they affect the parameters? Where do I look? > > Martin > > > On 4 June 2014 10:18, Anders Logg <[email protected]> wrote: > > On Wed, Jun 04, 2014 at 09:32:24AM +0200, Martin Sandve Alnæs wrote: > > The next buildbot fails with a compile error in bench because of some > changes > > to Assembler last night. > > > > Before that it failed because of > > > > > > ERROR: test_nasty_jit_caching_bug (__main__.JIT) > > ---------------------------------------------------------------------- > > Traceback (most recent call last): > > File "./test.py", line 36, in test_nasty_jit_caching_bug > > parameters["form_compiler"]["representation"] = representation > > File "/home/buildbot/fenicsbbot/next/dolfin-full/lib/python2.7/ > site-packages/ > > dolfin/cpp/common.py", line 2087, in __getitem__ > > raise KeyError, "'%s'"%key > > KeyError: "'form_compiler'" > > > > > > and then a segfault. I don't think those are mine? > > Your changesets seem to be the only ones that affect parameters in the > Python layer, so it's likely you... > > > > > Martin > > > > > > > > On 3 June 2014 21:59, Anders Logg <[email protected]> wrote: > > > > > > On Tue, Jun 03, 2014 at 09:58:41PM +0200, Martin Sandve Alnæs wrote: > > > > > > > Exactly why I volunteered to reset to next for Johannes ;) > > > > > > I'll be more suspicious next time.... ;-) > > > > > > > If you have any questions about the merge just say so. > > > > > > Seems to work fine with some minor melding. > > > > > > > > > > > > > Martin > > > > > > > > 3. juni 2014 21:11 skrev "Anders Logg" <[email protected]> følgende: > > > > > > > > On Tue, Jun 03, 2014 at 08:08:34PM +0100, Garth N. Wells wrote: > > > > > > > > > > On Tue, 3 Jun, 2014 at 7:59 PM, Anders Logg <[email protected]> > wrote: > > > > > >I'm working on a topic branch that I want to merge into > `next` > and > > > > > >then > > > > > >`master` (for ffc). > > > > > > > > > > > >I first merged `master` into my branch and fixed some issues > > > > > >Then I tried to merge my branch into `next`. > > > > > > > > > > > >I believed that any problems would be discovered in the first > step > > > > > >(merging from `master`) but the problems show up when I merge > into > > > > > >`next`. Is this expected? > > > > > > > > > > > >Or is it just that someone happened to merge something > conflicting > > > > > >with my changes into `next` which is not yet in `master`? > > > > > > > > > > Possibly. I'm also having unexpected trouble merging into > next. > Was > > > > > 'next' properly reset to 'master? > > > > > > > > Looking more closely at it, it seems the "problem" is some work > that > > > > Martin merged into `next` just after the reset. > > > > > > > > So the lesson is to merge quickly into `next` after the reset to > get > > > > first... :-) > > > > > > > > > > _______________________________________________ fenics mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics
