2.0.1 Release

[gassmoeller]

During the last two weeks we found a few bugs in the 2.0.0 release that might make it necessary to do a 2.0.1, because they cause crashes of valid input files, or silently produce wrong results. Specifically, I am thinking of the following:

• Fix bug described in multithreading for DG assemblers #2258
• Fix base model initialization for depth dependent material model #2253
• the bug in heating_model/compositional_heating that is fixed in Unify compute volume fractions #2199 (we might need a different smaller fix for the old version, or just pull the whole thing into 2.0.1, no functional change, just refactoring)
• the fix to the bug reported in track the location of the rising stokes sphere #2243 (yet to be implemented in ASPECT, part in Fix extract_used_vertices for calls with distributed MappingQEulerian dealii/dealii#6630)
• another bug that prevents combining particles and free surface in ASPECT 2.0.0 (I need to open an issue and PR for that)

We might also decide to just drop support for combining particles and free surface (the last two issues) by explicitly disabling that for the moment (I have not tested it, it is just not explicitly disabled).

Should we wait for a few weeks and do a 2.0.1 before the hackathon?

[jdannberg]

Yes, I think that makes sense.

[ljhwang]

Second. Lorraine Hwang Sent from my iPhone

…

On May 21, 2018, at 5:49 PM, Juliane Dannberg ***@***.***> wrote: Yes, I think that makes sense. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[bangerth]

Is anyone likely to run into any of these bugs? Things that just fail with an error are annoying, but at least they lead to an error one can report and be told to use the dev sources instead. (Though they may scare people away from using ASPECT.) The ones that silently produce wrong results concern me more. #2199 is of that sort.

The alternative is to stabilize for a month after the hackathon and then put out a 2.1. The people who go to the hackathon will all use the dev sources anyway, so they're not a concern. I do wonder, however, whether we're going to be able to stabilize soon after the hackathon. We've usually had a very long tail of things from the hackathon...

[jdannberg]

@tjhei, @gassmoeller and I just discussed this and agreed to make a 2.0.1 release (that only includes the bugfixes cherry-picked onto a new branch) at the hackathon.

[tjhei]

that is fixed in #2199 (we might need a different smaller fix for the old version, or just pull the whole thing into 2.0.1

I think we can pull the whole thing...

[gassmoeller]

We should now have everything that is necessary for 2.0.1. Lets move forward with it.

[tjhei]

clang 6 with deal.II 9.1.0-pre fails with:

CMakeFiles/aspect.dir/source/mesh_refinement/temperature.cc.o:temperature.cc:function aspect::MeshRefinement::Temperature<3>::execute(dealii::Vector<float>&) const: error: undefined reference to 'void dealii::KellyErrorEstimator<3, 3>::estimate<dealii::TrilinosWrappers::MPI::BlockVector, dealii::DoFHandler<3, 3> >(dealii::Mapping<3, 3> const&, dealii::DoFHandler<3, 3> const&, dealii::Quadrature<2> const&, dealii::FunctionMap<3, dealii::TrilinosWrappers::MPI::BlockVector::value_type>::type const&, dealii::TrilinosWrappers::MPI::BlockVector const&, dealii::Vector<float>&, dealii::ComponentMask const&, dealii::Function<3, double> const*, unsigned int, unsigned int, unsigned int, dealii::KellyErrorEstimator<3, 3>::Strategy)'

(and a couple other). @bangerth have you seen that?

[tjhei]

Furthermore, I am getting deprecation warnings with 9.1.0-pre of course (ConstraintMatrix, tria_boundary_lib). Should we try to suppress those for 2.0.1?

[gassmoeller]

the tria_boundary_lib would be easy to fix by cherry picking #2292 and #2329, I prepare another one to fixup #2329.

[bangerth]

The linker errors are news to me.

[tjhei]

Okay, release got tagged. Left to do:

• announce
• update cig website
• update bib entries for zenodo, etc.

[tjhei]

@gassmoeller do you want to send the announcement (see release tasks)?

[gassmoeller]

Ok, announcement is out. The zenodo entries on the CIG website and on our website seem up-to-date as well. The only thing left would be a new figshare upload for the manual? Or is there something else?

[tjhei]

Thanks. The manual is identical except the version number on the title. I think @bangerth was proposing to ignore this and not upload a new file (and in the future put 2.0 instead of 2.0.0 as the title). I can upload it but then we need to update all the bib entries...

[gassmoeller]

Lets keep it as is then.