Updated taskmaster to check srcnode for variant dir files without builders

[dmoody256]

fixes #2908

https://discord.com/channels/571796279483564041/971831743944491008/976472063990718494

Contributor Checklist:

• I have created a new test or updated the unit tests to cover the new/changed functionality.
• I have updated CHANGES.txt (and read the README.rst)
• I have updated the appropriate documentation

[mwichmann]

CI failed the "Linux Testing" build with an oddity - I've never seen F77PATH.py fail. The test log isn't terribly informative, but seems to be building something when no build was expected - the test line (#152) is:

test.up_to_date(arguments = args)

and from the log of what actually ran there was:

...
> scons: `variant/prog' is up to date.
> f77 -o subdir/prog.o -c -x f77 -Isubdir/include -Isubdir -Isubdir subdir/prog.f77
> gfortran -o subdir/prog subdir/prog.o
> scons: done building targets.

[dmoody256]

Yeah its weird because when I run the test with manual commands, I can't reproduce it.

Been looking into it, but because I can not reproduce it manually, I am force to debug through the test which is not very productive.

When I added --debug=explain, it said rebuilding because it doesn't exist, that's where I am at now.

[mwichmann]

Yes, that's a real pain. @bdbaddog has in the past tried to suggest some techniques for interacting with the CI, which I've never gotten to work.

[mwichmann]

Do you have an f77 command? I don't.... I'm not used to the command even running. Github isn't as clear on how the images are provisioned as AppVeyor is (or as Travis was).

[dmoody256]

Do you have an f77 command? I don't.... I'm not used to the command even running. Github isn't as clear on how the images are provisioned as AppVeyor is (or as Travis was).

I should be more clear, I can reproduce this via the test on a local machine, which is much better than needing to debug through CI. However I am perplexed how me manually running the same commands works, but the test running them fails.

[mwichmann]

Usually, that means os.environ != env['ENV']... wonder if there's some important env var not picked up. Where does your f77 come from? Oh, wait, it looks like Ubuntu provides an f77 symlink where Fedora doesn't, so that's where it comes from.

Environment vars sound like they're a red herring:

The gfortran compiler currently does not make use of any environment variables to control its operation above and beyond those that affect the operation of gcc.