Fix/git detached head

[pgierz]

Should fix the broken git branch stuff with detached head, sorry about that.

[pgierz]

There is probably a better way to do this, but my brain is fried today. Someone please double check.

[mandresm]

Some models are still broken in the tests. I'll review this tomorrow and maybe even have a look at the models having problems (AWICM3)

[pgierz]

As with the doc-test branch, all failing tests are now from esm-tests, and all seem to be "Run Files Differ"

[mandresm]

#bump

[pgierz]

Will close #734, still needs some manual testing

[pgierz]

Manually tested with AWICM3 frontiers-xios and AWICM3 3.1.

@JanStreffing: something is still weird. But if you are happy with this, then that is how it shall be. Currently xios does not seem to have a model directory, and rnfmap is not a git-controlled thing.:

$ cat ../experiments/git_test_001/run_20000101-20011231/log/git_test_001_vcs_info.yaml

esm_tools: d3e90a56
fesom:
  branch_name: awicm-3-frontiers_parallel-restart
  diffs:
  - ''
  hash: 9b27593c
  path: /work/ab0995/a270077/SciComp/Model_Support/AWICM3/broken_git_check/model_codes/awicm3-frontiers-xios///fesom-2.0
oasis3mct:
  branch_name: awicm-3-frontiers
  diffs:
  - ''
  hash: 347e788
  path: /work/ab0995/a270077/SciComp/Model_Support/AWICM3/broken_git_check/model_codes/awicm3-frontiers-xios///oasis
oifs:
  branch_name: awicm-3-frontiers-xios
  diffs:
  - diff --git a/src/ifs/module/yommcc.F90 b/src/ifs/module/yommcc.F90
  - index e7d74a1..131227a 100644
  - '--- a/src/ifs/module/yommcc.F90'
  - +++ b/src/ifs/module/yommcc.F90
  - '@@ -180,7 +180,7 @@ INTEGER(KIND=JPIM) :: NFRCPL'
  - ' '
  - ' !Logical switches for Coupling'
  - ' LOGICAL :: COUPLENEMOECE = .FALSE.   ! Use EC-Earth NEMO coupling'
  - '-LOGICAL :: COUPLEFESOM2 = .FALSE.    ! Use AWI FESOM2 coupling'
  - '+LOGICAL :: COUPLEFESOM2 = .TRUE.    ! Use AWI FESOM2 coupling'
  - ' LOGICAL :: COUPLENEMOFOCI = .FALSE.   ! Use FOCI NEMO coupling'
  - ' LOGICAL :: LNEMOLIMGET = .FALSE.   ! Get new fields'
  - ' LOGICAL :: LNEMOLIMPUT = .FALSE.   ! Put new fields'
  hash: c78fa46
  path: /work/ab0995/a270077/SciComp/Model_Support/AWICM3/broken_git_check/model_codes/awicm3-frontiers-xios///oifs-43r3
rnfmap: Not a git-controlled model!
xios: Unable to locate model_dir for xios.

...and, yes, I am not particularly happy that diffs is a list of strings. I might still change that.

[JanStreffing]

Not sure why it's saying rnfmap is not a git controlled thing. It certainly is:
https://gitlab.dkrz.de/ec-earth/runoff-mapper
XIOS seems to use the model_dir from awicm3 and install_bins: bin/xios.x
No harm in adding a model_dir for consistencies sake I guess. Though it seems required to run the model.

[pgierz]

Jan, have a look in levante: /work/ab0995/a270077/SciComp/Model_Support/AWICM3/broken_git_check/model_codes/awicm3-frontiers-xios

It might be that the install_bins part is generating folders for things that aren't git repos (or at least, are not git repos where the model_dir is defined). That's too deep in the AWI-CM3 config for me to be comfortable touching it though...

[mandresm]

There seems to be a problem in FOCI (blogin):

    File "/github/home/.local/bin/esm_runscripts", line 11, in <module>
      load_entry_point('esm-tools', 'console_scripts', 'esm_runscripts')()
    File "/__w/esm_tools/esm_tools/src/esm_runscripts/cli.py", line 278, in main
      Setup()
    File "/__w/esm_tools/esm_tools/src/esm_runscripts/sim_objects.py", line 69, in __call__
      self.prepcompute()
    File "/__w/esm_tools/esm_tools/src/esm_runscripts/sim_objects.py", line 161, in prepcompute
      self.config = prepcompute.run_job(self.config)
    File "/__w/esm_tools/esm_tools/src/esm_runscripts/prepcompute.py", line 36, in run_job
      config = evaluate(config, "prepcompute", "prepcompute_recipe")
    File "/__w/esm_tools/esm_tools/src/esm_runscripts/helpers.py", line 68, in evaluate
      config = esm_plugin_manager.work_through_recipe(
    File "/__w/esm_tools/esm_tools/src/esm_plugin_manager/esm_plugin_manager.py", line 141, in work_through_recipe
      config = getattr(submodule, workitem)(config)
    File "/__w/esm_tools/esm_tools/src/esm_runscripts/prepare.py", line 707, in add_vcs_info
      if helpers.is_git_repo(model_dir):
    File "/__w/esm_tools/esm_tools/src/esm_runscripts/helpers.py", line 329, in is_git_repo
      git.Repo(path).git_dir
    File "/github/home/.local/lib/python3.8/site-packages/git/repo/base.py", line 152, in __init__
      raise NoSuchPathError(epath)
  git.exc.NoSuchPathError: /__w/esm_tools/esm_tools/src/esm_tests/resources/automatic_testing/comp/focioifs/focioifs-2.0/oifs-43r3
  Copying standard yamls from:  /__w/esm_tools/esm_tools/configs/
  Copying standard namelists from:  /__w/esm_tools/esm_tools/namelists/
  /__w/esm_tools/esm_tools/src/esm_tests/resources/runscripts/focioifs/focioifs-piCtl-test_lowcpu.yaml

Do you want me to try to reproduce it in a real computer?

[pgierz]

No, that's something else, and actually is still a bug in the git part. I'll look after I do the pool stuff.

[pgierz]

Broken tests are at this point "wrong truth". The technical elements seem to working fine. The error reported by @mandresm is now fixed. I am not 100% sure if my fix is the best way of doing it.

The offending function checks: is this a git directory? You get True if it is a git dir (makes sense), and you get False if it is not (make sense) OR False if it is non-existent (this part I am not sure about). Throwing the FileNotFound error might be better here, but that kills our tests. That FileNotFound case only happens on CI; so it will never be a problem in real production.

In any case, I would vote: please merge. :-)

[mandresm]

The offending function checks: is this a git directory? You get True if it is a git dir (makes sense), and you get False if it is not (make sense) OR False if it is non-existent (this part I am not sure about). Throwing the FileNotFound error might be better here, but that kills our tests. That FileNotFound case only happens on CI; so it will never be a problem in real production.

Agreed, this is the sort of thing that is going to be tested in the actual esm_tests, and we have to run those from time to time anyway, so I think it's okay if it doesn't work for check tests.

[mandresm]

Great feature, thanks for fixing it! :)

[mandresm]

updating the truth...

[mandresm]

All tests passed (except AWIESM icebergs due to lack of update related to an ongoing issue), #bump!