GitLab

Created by: bartlettroscoe

One (or more) of the commits (or the combination of commits) pushed by several people broke one Intrepid2 test and four MueLu tests in this CI build on 5/4/2017:

• http://testing.sandia.gov/cdash/index.php?project=Trilinos&parentid=2876438&filtercount=3&showfilters=1&field1=groupname&compare1=61&value1=Continuous&field2=buildstarttime&compare2=84&value2=now&filtercombine=and

which shows the failing tests:

• Intrepid2_refactor_unit-test_Shared_Polylib_Serial_Test_01_MPI_1
• MueLu_UnitTestsEpetra_MPI_1
• MueLu_UnitTestsEpetra_MPI_4
• MueLu_UnitTestsTpetra_MPI_1
• MueLu_UnitTestsTpetra_MPI_4

The commits that were pulled in the following CI iteration on 5/4/2018 involve the range of Trilinos commits 75ba6dfe..1d4823c5 and the commits are shown here:

• http://testing.sandia.gov/cdash/viewNotes.php?buildid=2876439##note1

The commits pulled in this CI iteration were by the people:

$ git shortlog -ns 75ba6df..1d4823c
     6  Kyungjoo Kim
     2  Nathan Ellingwood
     1  Alejandro Mota
     1  Dan Ibanez
     1  Mark Hoemmen

The commits showing the standard CI build was run are:

$ git log-short --grep="Build/Test Cases Summary" 75ba6df..1d4823c
26d683b "MiniTensor #1264: Switch from Teuchos::ScalarTraits to Kokkos::Detail::ArithTraits."
Author: Alejandro Mota <amota@sandia.gov>
Date:   Thu May 4 13:22:29 2017 -0700 (4 days ago)

Therefore, any of these other 4 individuals may have broken the standard CI build. One would have to examine the individual commits and do more investigation to figure our what changes are breaking these tests.

Also, note that these tests are failing for all ALL the "Nightly" and (ironically) "Clean" builds for MueLu as shown at:

• http://testing.sandia.gov/cdash/index.php?project=Trilinos&date=2017-05-07&subproject=MueLu

The only build that does not show MueLu test failures is the Epetra-only build (Linux-GCC-5.2.0-SERIAL_DEBUG_DEV_MueLu_Epetra). Therefore, this is a wide-spread failure.

NOTE: I am not mentioning anyone or adding any labels yet because I want to see how long it takes for the current Trilinos development community to discover and fix these failures. I am just creating this issue so that I can use the Issue ID number in the commit where I disable these four tests in pre-push testing (but not post-push testing). This is needed so that we don't stop people trying to use the checkin-test-sems.sh script to push. We can't have recuring failures being ignored from stopping other people from using the checkin-test-sems.sh script.