Setting up automated tests for optional third party library

[rks171]

I have created an optional dependency on a third party library (TPL) in the MOOSE app I am working with. The way in which I did this is:

1. Compile the TPL separate from my MOOSE app but using the same env as was used for MOOSE compilation
2. Modify the MOOSE app Makefile to add the TPL lib/include dirs to the ADDITIONAL_INCLUDES and ADDITIONAL_LIBS variables and also define the TPL name in ADDITIONAL_CPPFLAGS.
3. Create a class in my MOOSE app for interfacing with this TPL. Ifdef statements are used to only include the TPL calls if the TPL variable was defined in ADDITIONAL_CPPFLAGS.

This allows me to use the TPL successfully as long as it was included in the build. Now I would like to add some automated tests using the MOOSE testrunner to protect this feature. However, I need for the test to only be run if the TPL was included during the build and ignored completely otherwise. Is there a way to do this?

[GiudGiud]

Hello

You need to set up a test group for automated testing. All the tests that will use the TPL will have to be in that group, and when you want to run them you will have to type:

./run_tests -g group_name_with_tpl

Add in the tests file for each concerned test:
group = 'group_name_with_tpl'

Guillaume

[rks171]

I tested moving these tests out of my yourapp/test dir and into the root directory, yourapp, and the runner still finds the tests, so it must just be searching everything from the root down.

[GiudGiud]

Umm. Well I understand you are trying to keep the standard behavior to be ./run_tests and not ./run_tests + some option but I m running out of ideas.
@loganharbour anything I missed to be able to exclude a group of tests by default?

[loganharbour]

I wouldn't suggest groups here. We should just add the dependency as a test spec option that we can check for. That way, @rks171 can do something like tpl=true in his spec.

[rks171]

@loganharbour, are you saying this would need to be added as a new feature to the test runner? And would it work by the caller passing in some flag to the run_tests script?

[loganharbour]

Yes, that's correct

[rks171]

I want to point out that there is another solution for this using submodules. If the TPL is pulled into the MOOSE app as a git submodule, then any tests that are defined that require that TPL can have the following input added to the test definition:

required_submodule = 'submodule name'

If the submodule is present, the tests will be run, and if not, they will be skipped.