BUGFIX: MinGW, platform-agnostic CI

[scivision]

• MinGW strftime didn't work, this fixes that by linking ucrtime for MinGW
• strftime('%s') isn't supported everywhere (e.g. Windows doesn't have it) so use datetime itself. NOTE: was this "dummy" variable necessary?

These changes make all tests pass on Windows MinGW/MSYS

CI move

Travis-ci.org is shutting down. Travis-CI.com now has quota even for open-source. Move CI to GitHub Actions (you need to enable this from this repo Actions tab).

Make CI be Windows, Mac, Linux to avoid platform-specific issues

lint / cleanup

• use, intrinsic just for reminder these are language standard

[milancurcic]

Thank you for these improvements and especially Windows and CI fixes.

I'm only not in favor of https://github.com/wavebitscientific/datetime-fortran/pull/70/commits/fb39ee972a51dec8de1c2c0d7df91304b53f373b.

As I understand it, the only pro is:

• Less verbose code, which I like.

And the cons are:

• F2008 feature, so older compiler versions may not support it. I don't know when this feature was implemented by compilers. I wrote most of the datetime in 2013 so it doesn't use too advanced features--it worked nicely with older compilers. I'd prefer not dropping that for a small benefit.
• While less verbose, the reader also needs to know and remember that rule themselves. I find that people tend to be more familiar with some older features (F77, F90) and less so with newer (F2003 onward).

So I suggest reverting this commit. Unless you can convince me otherwise, I'm quite open to convincing :).

[milancurcic]

Here's another reason why I still like writing pure elemental even when pure is redundant with elemental:

I consider pure Fortran procedures to be overall beneficial for any code base. They help make cleaner code, and may even help compilers optimize.

When I write pure elemental, I'm also signaling and promoting the use of pure wherever possible. When you write just elemental, yes, they're still pure, but it's implied. The reader needs to be aware of this rule. When you write pure elemental, the word pure is right there in your face.

So it's kind of a hand-wavy, persuasion-based argument, but I thought it worth mentioning.

[scivision]

the "pure elemental" change is really bikeshedding or like 2 vs 4 spaces ha ha. I dropped that commit

[scivision]

So to make the new CI work, please go to the Actions tab at the top of this repo and enable. It might take one more commit to the *.f90 code to make the CI trigger then.

[milancurcic]

Great. I went to Actions but I don't see a simple "Enable" button or something similar. Wasn't sure what to do so I added you as a repo maintainer. Do you mind doing it? :)

I tested this PR with GNU and Intel compilers on Linux and all looks good on my end. I trust you with the Windows stuff.

[scivision]

OK thanks. They've changed the interface, I will figure it out. I was an early Actions user and I enabled it once and i'll see what it needs now.

[scivision]

OK I had to make a dummy PR enabling Actions which I then closed. Now Actions is enabled.

[milancurcic]

Please merge when you're satisfied with this. I am. Thanks again!

[scivision]

the "needs" parameters in the CI.yml are just to be a good citizen--they avoid wasting MacOS and Windows workers, by waiting to be sure the "cheaper" Linux CI worker passes first. I do this even if it's all Linux workers, avoid needlessly failing a whole bunch of CI by doing sanity checks first.

[milancurcic]

I'm confused how did my commit from yesterday not make it into this PR. I pushed it yesterday:

commit 2d02318728bbba5de51450f5d73bd8cb311a7a9d (HEAD -> redun, origin/redun)
Author: milancurcic <caomaco@gmail.com>
Date:   Thu Dec 17 10:05:07 2020 -0500

    bump version to 1.7.1

@scivision do you know?

[scivision]

I rebased and our commits may have overlapped in time. I.e. I erased the commit unknowingly

[scivision]

So I think just recreate the commit and push again? Hopefully it was just a couple lines..

[milancurcic]

Yes, it's an easy and small commit. I will send it soon.

I'm more shocked that git allowed this. My commit has disappeared without a trace. But then my git knowledge is limited. Maybe this is what rewriting commit history does.