Create v1.4.0 release #2029

jwpeterson · 2019-02-10T00:10:08Z

Reference this ticket in bug fixes, etc. that get cherry-picked over from master onto the release branch.

Refs #2029.

The Elem::set_p_level() API is only available when LIBMESH_ENABLE_AMR, since it is a feature used for p-refinement. Refs #2029.

LIBMESH_CHKERR must be called from a ParallelObject when exceptions are not enabled. That is because, in that specific case, the LIBMESH_CHKERR macro just forwards on to the old CHKERRABORT macro, which requires a communicator. Refs #2029.

Therefore we should not have a unit test with a std::vector<Node>. Refs #2029.

Refs #2029.

I think because of the update to automake 1.16.1, build-aux/config.guess is now reporting that my system is "x86_64-pc-linux-gnu", so I needed to add a new symbolic for the Tecplot binaries. Note that you have to force add (add -f) these links because otherwise they are ignored by the "lib" line in .gitignore. Refs #2029.

I thought about having "make distcheck" automatically initialize/update the metaphysicl submodule, but that's way too complicated/too many things can go wrong so I settled on just throwing a human-readable error if the user runs "make dist" or "make distcheck" without first initializing the submodule. Refs #2029.

Refs #2029.

The basic idea is that we don't want to mix building our own METIS and using PETSc's ParMETIS. The different possibilities are: | Parmetis in PETSc? | yes | no ---------------------- yes | A | B METIS in PETSc? ---------------------- no | C | D A: PETSc was built with both METIS and ParMETIS support. In this case, we use both from PETSc. B: PETSc is built with METIS support but no ParMETIS support. In this case, we use the METIS support and disable ParMETIS. C: PETSc is built with no METIS support, but does have ParMETIS. I'm not sure this is really possible, but in this case, we would build our own METIS and disable ParMETIS support just to avoid mixing them. D: PETSc is built with neither METIS nor ParMETIS support. In this case, we build both on our own. Refs #2029.

This doesn't change any code, it just updates some of the bootstrapped files to use the same version of automake as libmesh so there are no diffs when we run bootstrap at the libmesh level. Refs #2029.

Refs #2029.

There was an issue with subdomain restricted variables. I previously did not realize that they would still show up in the discontinuous solution vector (as zero padding) so I was not indexing correctly in that case. This fix should also be cherry-picked onto v1.4.0 since this API is present in that version. Refs libMesh#2029. Refs libMesh#2010.

Instead of setting TAR_OPTIONS, I also tried calling AC_SUBST(am__tar,[tar --hard-dereference]) but unforunately this alternate approach did not work, perhaps the AC_SUBST comes too late in the process to take effect or perhaps it is simply not possible to do this. It was an off-hand remark in a mailing list [0], not an officially sanctioned way of doing things. [0]: http://gnu-automake.7480.n7.nabble.com/bug-19616-dist-tarball-contains-hardlinks-td21667.html Refs libMesh#2046. Refs libMesh#2029.

There was an issue with subdomain restricted variables. I previously did not realize that they would still show up in the discontinuous solution vector (as zero padding) so I was not indexing correctly in that case. This fix should also be cherry-picked onto v1.4.0 since this API is present in that version. Refs #2029. Refs #2010.

Instead of setting TAR_OPTIONS, I also tried calling AC_SUBST(am__tar,[tar --hard-dereference]) but unforunately this alternate approach did not work, perhaps the AC_SUBST comes too late in the process to take effect or perhaps it is simply not possible to do this. It was an off-hand remark in a mailing list [0], not an officially sanctioned way of doing things. [0]: http://gnu-automake.7480.n7.nabble.com/bug-19616-dist-tarball-contains-hardlinks-td21667.html Refs #2046. Refs #2029.

Currently if not LIBMESH_HAVE_CXX14_MAKE_UNIQUE we define libmesh_make_unique to be equivalent to make_unique without namespace scoping. I believe explicitly scoping to libMesh::make_unique makes sense Refs #2029.

Refs #2029.

jwpeterson · 2019-03-15T18:51:29Z

1.4.0 was released on March 11, 2019. There have been a couple of small bug fixes since then, so I'm contemplating a 1.4.1 release at some point if enough of them accumulate. I'll keep using this issue number for those bug fixes.

Really we should be rejecting 1.16 too now but that's a less trivial fix Refs #2029.

Otherwise debug-mode PETSc builds (rightly) scream at us for starting set operations without finishing add operations. Refs #2029.

Refs #2029.

Fixes one of the issues reported in #2071. Note: our current CIVET complex testing is not running or compiling the unit tests, so this slipped through the cracks. I have a different PR in that will hopefully address the unit testing issue so this won't happen again. Refs #2029.

This API was present in 1.4.0, so this commit should also be cherry-picked onto the in-progress v1.4.1_branch. Refs libMesh#2029.

This API was present in 1.4.0, so this commit should also be cherry-picked onto the in-progress v1.4.1_branch. Refs #2029.

Refs #2029. Closes #2087.

This function did not wind up making it into any PETSc release. This should probably also be cherry-picked onto the v1.4.1 branch. Even though the likelihood of any tagged libmesh release continuing to work with PETSc master is negligibly small, the tagged release _might_ work with a future PETSc 3.11 so we should try and support that if possible. Refs #2084. Refs #2029.

This would have been found by "make installcheck" on a build with CapnProto, SLEPC, and GLPK enabled, if we ever ran such a thing... This commit will also be cherry-picked onto the 1.4.1 branch. Refs #2029.

We've never been correctly copying remote_elem neighbor links in the skip_find_neighbors code path (or rather, we've been copying them correctly and then accidentally *undoing* them), and since #2028 made skip_find_neighbors the default behavior in more code paths, that bug was starting to be triggered e.g. in the #1938 tests. This commit gets remote_elem into the old-to-new-elements map so the skip_find_neighbors code path will recognize it, and removes a bit of redundancy in the rest of the old-to-new-elements map initialization, and moves the manual remote_elem-specific copying code into the !skip_find_neighbors path to avoid redundancy with that. Refs #2029.

This was part of a half-baked attempt to support both "refine and coarsen prior to EquationSystems::reinit for CPU efficiency" and "refine and coarsen only within EquationSystems::reinit for memory efficiency" idioms, but the latter support was seriously broken. Refs #2029.

This is a little more readable Backported with the other commits from #2099 with the command. git cherry-pick e9297f1 33466d0 ae5cbc1 de07ffc Refs #2029.

This catches a bug much earlier in runtime: when bad constraint equations are created, rather than when they are used. Backported with the other commits from #2099 with the command. git cherry-pick e9297f1 33466d0 ae5cbc1 de07ffc Refs #2029.

This turned out to be a red herring in my search for a different bug, but it's not a bad thing to keep double-checking. Backported with the other commits from #2099 with the command. git cherry-pick e9297f1 33466d0 ae5cbc1 de07ffc Refs #2029.

If we have a subdomain-restricted variable which is supported on a fine element but *not* supported on a neighboring coarse element, then we do *not* want to try to generate hanging node constraints for that interface. Doing so was giving us OOB reads and garbage DoF IDs. Backported with the other commits from #2099 with the command. git cherry-pick e9297f1 33466d0 ae5cbc1 de07ffc Refs #2029.

We only set h_min in this way if it has not been previously updated. This ensures that h_min is the min over both meshes being stitched. Refs #2029. Co-authored-by: John W. Peterson <[email protected]>

Refs #2029.

Refs #2104. Refs #2115. Refs #2029.

Otherwise we'll happily set inconsistent flags and not discover the error until we actually start trying mesh refinement. Refs #2029.

Now that we're asserting error indicator sanity, we need to be sure that our linear solvers really converge even when no special LIBMESH_OPTIONS are used. Refs #2029.

jwpeterson added a commit that referenced this issue Feb 10, 2019

Update automake version number to 1.4.0-rc1.

5dc3512

Refs #2029.

jwpeterson added a commit that referenced this issue Feb 10, 2019

Run bootstrap.

1887330

Refs #2029.

jwpeterson added a commit that referenced this issue Feb 10, 2019

Update NEWS file for 1.4.0 release.

a501acf

Refs #2029.

jwpeterson mentioned this issue Feb 11, 2019

Add shape and skew metrics for Hexes. #2030

Merged

jwpeterson added a commit that referenced this issue Feb 11, 2019

Skip unit test when !LIBMESH_ENABLE_AMR.

c58c9fe

The Elem::set_p_level() API is only available when LIBMESH_ENABLE_AMR, since it is a feature used for p-refinement. Refs #2029.

jwpeterson added a commit that referenced this issue Feb 11, 2019

Node copy constructor is deprecated.

4b10e9f

Therefore we should not have a unit test with a std::vector<Node>. Refs #2029.

jwpeterson mentioned this issue Feb 11, 2019

Fixed petsc_linear_solver for PETSc-master #2027

Merged

jwpeterson pushed a commit that referenced this issue Feb 11, 2019

Fixed petsc_linear_solver for PETSc-master

4004a24

Refs #2029.

This was referenced Feb 15, 2019

Several fixes to get "make distcheck" working again #2037

Merged

Fix issue with PETSc that has METIS but no ParMETIS #2036

Merged

Update metaphysicl submodule. #2038

Merged

jwpeterson added a commit that referenced this issue Feb 16, 2019

distcheck: clean up unit test output files.

b77ae84

Refs #2029.

jwpeterson added a commit that referenced this issue Feb 16, 2019

Run bootstrap.

1517416

Refs #2029.

jwpeterson mentioned this issue Feb 21, 2019

ExodusII_IO: Fix write_element_data_from_discontinuous_nodal_data() #2045

Merged

jwpeterson mentioned this issue Feb 21, 2019

Set TAR_OPTIONS to improve portability of tarballs. #2048

Merged

This was referenced Feb 26, 2019

ExodusII_IO: Drop inactive variables from the lists passed to the Exodus helper. #2052

Merged

Explicitly specify scope in else branch: libmesh_make_unique macro #2053

Merged

jwpeterson pushed a commit that referenced this issue Feb 27, 2019

fix unique_id update

1713968

Refs #2029.

jwpeterson added a commit that referenced this issue Mar 1, 2019

Update automake version number to 1.4.0.

dbbfeb3

Refs #2029.

jwpeterson added a commit that referenced this issue Mar 1, 2019

Run bootstrap.

8e63486

Refs #2029.

jwpeterson pushed a commit that referenced this issue Mar 15, 2019

Correctly identify automake 1.15 as too old

25c4d17

Really we should be rejecting 1.16 too now but that's a less trivial fix Refs #2029.

jwpeterson pushed a commit that referenced this issue Mar 15, 2019

Close residual/vector before applying constraints

1546d48

Otherwise debug-mode PETSc builds (rightly) scream at us for starting set operations without finishing add operations. Refs #2029.

jwpeterson added a commit that referenced this issue Mar 15, 2019

Update automake version number to 1.4.1.

947b570

Refs #2029.

jwpeterson added a commit that referenced this issue Mar 15, 2019

Run bootstrap.

9e6dea3

Refs #2029.

jwpeterson pushed a commit that referenced this issue Mar 15, 2019

Capture iterations from PETSc in diff solver

1f2d962

Refs #2029.

jwpeterson added a commit that referenced this issue Mar 15, 2019

Update NEWS file for potential 1.4.1 release.

ecdc14b

Refs #2029.

jwpeterson added a commit to jwpeterson/libmesh that referenced this issue Mar 18, 2019

FEInterface::get_continuity() should take a const reference.

c59d723

This API was present in 1.4.0, so this commit should also be cherry-picked onto the in-progress v1.4.1_branch. Refs libMesh#2029.

jwpeterson mentioned this issue Mar 18, 2019

FEInterface::get_continuity() should take a const reference. #2076

Merged

jwpeterson added a commit that referenced this issue Mar 18, 2019

FEInterface::get_continuity() should take a const reference.

87274c1

This API was present in 1.4.0, so this commit should also be cherry-picked onto the in-progress v1.4.1_branch. Refs #2029.

jwpeterson mentioned this issue Mar 26, 2019

VecScatterCreateWithData is not happening #2084

Closed

jwpeterson added a commit that referenced this issue Mar 26, 2019

Fix doco in DenseMatrix::right_multiply.

9769b2b

Refs #2029. Closes #2087.

This was referenced Apr 10, 2019

Fix 2006 bug with refinement within ES::reinit() #2100

Merged

Subdomain-restricted variables hanging nodes fix #2099

Merged

jwpeterson pushed a commit that referenced this issue Apr 10, 2019

Refactor to reduce indentation

ef212cd

This is a little more readable Backported with the other commits from #2099 with the command. git cherry-pick e9297f1 33466d0 ae5cbc1 de07ffc Refs #2029.

jwpeterson mentioned this issue Apr 17, 2019

Mesh stitching fixes for NodeElems #2103

Merged

jwpeterson added a commit that referenced this issue Apr 18, 2019

Fall back on absolute distance check in all-NodeElem case.

ad6e667

Refs #2029.

jwpeterson added a commit that referenced this issue Apr 30, 2019

Remove deprecated Elem API calls.

c81c083

Refs #2104. Refs #2115. Refs #2029.

jwpeterson pushed a commit that referenced this issue May 2, 2019

Assert consistent ErrorVector values for flagging

cad1e84

Otherwise we'll happily set inconsistent flags and not discover the error until we actually start trying mesh refinement. Refs #2029.

jwpeterson pushed a commit that referenced this issue May 2, 2019

Restrict adaptivity ex3 max_r_steps

0fd745e

Now that we're asserting error indicator sanity, we need to be sure that our linear solvers really converge even when no special LIBMESH_OPTIONS are used. Refs #2029.

jwpeterson closed this as completed Sep 3, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Create v1.4.0 release #2029

Create v1.4.0 release #2029

jwpeterson commented Feb 10, 2019

jwpeterson commented Mar 15, 2019

Create v1.4.0 release #2029

Create v1.4.0 release #2029

Comments

jwpeterson commented Feb 10, 2019

jwpeterson commented Mar 15, 2019