we're on python3, so importing from future should no longer be necessary

[pavelkomarov]

Just a small PR. I might PR this repo more as I use it.

[pavelkomarov]

A lot of the fortran codes are .f90 files, yet here we're using fc.compiler_f77. That won't cause any issues? Is a fortran 1990 compiler available that we should be using instead?

[mandli]

I think for legacy support fc.compiler_f77 defaults to gfortran, which can handle both formats, usually anyway. I am not even sure that f77 compilers exist on anything modern anyway.

[mandli]

This is failing because of the shallow water test, which I think was failing before.

[pavelkomarov]

How can we fix that test? Should we replace mention of '77 with a function named for gfortran? I bet one exists.

[mandli]

I think the compiler mentioned should be fine and I am not certain what consequences it would have to change to a f90 version.

@ketch do you remember why that test is failing?

[pavelkomarov]

According to the numpy docs, distutils is legacy and may be removed in a future version https://numpy.org/doc/stable/f2py/buildtools/distutils.html

[ketch]

Yes, we can certainly remove those imports. But I think it would be best to do it all in one PR; otherwise we will clutter the history with dozen's of PRs just for this tiny change.

[ketch]

The failing test is discussed here: #642. Let's keep the discussion of it there so it's easier to follow.

[pavelkomarov]

Tried to squash commits and created three commits.

[ketch]

Thanks, that's great!

It looks like this is failing, and I'm guessing it's related to an absolute vs. relative import issue, but I don't remember what actually changes by not importing absolute_import. The actual error occurs during the build stage and is

../../pyclaw/src/petclaw/meson.build:37:5: ERROR: File cfl.py does not exist.

The file cfl.py does exist in that directory. I don't think the issue is with that file specifically (it is the first file, alphabetically, in the first subdirectory, alphabetically). Rather, this seems to have made it so meson can't find any of the source files.

[pavelkomarov]

Okay, any idea how I fix that? I don't think meson should be relying on importing from future. Do we need to do that import .package relative import kinds of constructions or something?

[ketch]

I spent a half hour looking into it but I still have no idea how to fix it.

[pavelkomarov]

If the build is this fragile, it will be difficult to make contributions.

[pavelkomarov]

Somehow I managed to delete a whole file there. Hopefully that was the problem.

[ketch]

Haha; of course that was the problem.