[AMBER-Developers] PSA: Please be careful when re-ordering Makefile targets

From: Daniel Roe <daniel.r.roe.gmail.com>
Date: Mon, 23 Mar 2020 14:45:20 -0400

Hi All,

I'm currently adding some new leap FF parameters, and after developing
a test for them I noticed that the test wasn't being run, even after
being added to the correct target. After some digging I figured out
this was because at some point in the past the targets in
$AMBERHOME/AmberTools/test/leap/Makefile had been reordered, and we
ended up with something like this (*** emphasis from me):

```

Test application of bond augmentation in advanced force fields.

diffing r1p7e.prmtop.save with r1p7e.prmtop
PASSED
==============================================================
*** make[3]: Nothing to be done for 'test.complete'. ***
make[3]: Leaving directory
'/iscratch/jenkins-cpu/workspace/AmberGitlab/AmberMergeGatePipeline/AmberTools/test/leap'
```

So effectively the test.complete target is being skipped. This is from
a recent CI build
(https://jenkins.jasonswails.com/blue/rest/organizations/jenkins/pipelines/AmberGitlab/pipelines/AmberMergeGatePipeline/runs/964/nodes/81/steps/207/log/?start=0)
- I haven't checked others but I'm pretty sure this has been happening
for a while. If I switch to a more "sane" ordering (i.e. targets only
reference other targets that have been previously listed) I get the
correct behavior, i.e. the tests are run.

The change in behavior appears to be related to commit
'94acd4a7515ed08fdc1...' from 2017, so apparently the tests have been
being skipped for *years*. And sure enough when I run those tests now
I get a bunch of failures - I'm trying to fix them now.

Bottom line: when you change around Makefiles, double check that you
get the behavior you want. Our test suite is far too monolithic to
assume everything is fine even after what should be an innocuous
change. I hope to have this specific error fixed by later
today/tomorrow, but as we do QA testing over the next month please
keep this in mind.

-Dan

_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Mar 23 2020 - 12:00:02 PDT
Custom Search