enable SLiM on Windows

[petrelharp]

I'm having a start at #1305 here.

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: -0.08 ⚠️

Comparison is base (5bb5b71) 99.85% compared to head (be9ff1c) 99.77%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1342      +/-   ##
==========================================
- Coverage   99.85%   99.77%   -0.08%     
==========================================
  Files         117      117              
  Lines        4032     4037       +5     
  Branches      566      565       -1     
==========================================
+ Hits         4026     4028       +2     
- Misses          3        5       +2     
- Partials        3        4       +1     

Impacted Files	Coverage Δ
stdpopsim/cli.py	99.10% <100.00%> (-0.68%)	⬇️
stdpopsim/slim_engine.py	100.00% <100.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[petrelharp]

There seems to be some issues with TemporaryDirectory in python 3.7 on Windows: https://bugs.python.org/issue26660 ... so, perhaps we will not support that combination?

[petrelharp]

Hmph, I'm getting inscrutible

E           stdpopsim.slim_engine.SLiMException: slim exited with code 3.

errors. Maybe someone can have a go at this on an actual windows box?

[petrelharp]

Looks like to get this totally working we'll need to fix MesserLab/SLiM#349 and build SLiM from source until a release happens.

[petrelharp]

Ben has commited a fix to MesserLab/SLiM#349; so we can now either build SLiM from source to run these tests or (this is my vote) wait for a release to enable this testing.

[petrelharp]

Tagging @andrewkern

[petrelharp]

However, @bhaller says it'd be nice to test SLiM on windows this way, so perhaps we should get this going?

[bhaller]

Mostly it'd be good to make sure no further Windows problems will bite stdpopsim after SLiM 4.0.1 rolls. It's infinitely easier to fix issues before rolling than to do a new roll just to fix an issue found post-roll. :-> I don't want to do a 4.0.2 unless the sky is falling. :->

[andrewkern]

i think we can build slim on windows simply enough like we do over on pyslim in a GH action- would that work?

[grahamgower]

Yeah, we used to build SLiM from sources in CI so this should be easy enough.

[bhaller]

i think we can build slim on windows simply enough like we do over on pyslim in a GH action- would that work?

SLiM builds itself on Windows in its GitHub Actions workflow, so you can look there if you hit any bumps in the road:

https://github.com/MesserLab/SLiM/blob/master/.github/workflows/tests.yml

[petrelharp]

I've restarted the tests because SLiM 4.0.1 is released, so that should fix the logflie problem.

[bhaller]

Still failed; looks like slim 4.0 might be hard-coded somewhere, @petrelharp?

2022-09-14T19:23:30.6130695Z   added / updated specs:
2022-09-14T19:23:30.6131434Z     - msprime==1.2.0
2022-09-14T19:23:30.6131904Z     - slim==4.0

4.0.1 is available on conda and should fix this problem.

[nspope]

Once #1343 is in this should be fixed ?

[petrelharp]

Hopefully I rebased this correctly - need to double-check some of those tests that maybe got removed.

[petrelharp]

Well, darn, I was hoping this was All Fixed, but it appears to not be fixed yet. I can't jump on this just now, but if anyone else is itching to do it, go for it (or remind me in a few weeks?)

[petrelharp]

For future reference, we're getting errors on Windows like:

_____________ TestAPI.test_recap_and_rescale_on_external_slim_run _____________
...
         out = re.sub(
            'defineConstant\\("trees_file.+;',
            f'defineConstant("trees_file", "{treefile}");',
>           out,
        )

tests\test_slim_engine.py:293: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
C:\Miniconda\envs\stdpopsim\lib\re.py:194: in sub
    return _compile(pattern, flags).sub(repl, string, count)
C:\Miniconda\envs\stdpopsim\lib\re.py:311: in _subx
    template = _compile_repl(template, pattern)
C:\Miniconda\envs\stdpopsim\lib\re.py:302: in _compile_repl
    return sre_parse.parse_template(repl, pattern)
...                    try:
                        this = chr(ESCAPES[this][1])
                    except KeyError:
                        if c in ASCIILETTERS:
>                           raise s.error('bad escape %s' % this, len(this))
E                           re.error: bad escape \U at position 32

C:\Miniconda\envs\stdpopsim\lib\sre_parse.py:1018: error

... but also

        if proc.returncode != 0:
            raise SLiMException(
>               f"{slim_path} exited with code {proc.returncode}.\n{stderr}"
            )
E           stdpopsim.slim_engine.SLiMException: slim exited with code 3.

stdpopsim\slim_engine.py:1685: SLiMException
------------------------------ Captured log call ------------------------------
WARNING  stdpopsim.cli:cli.py:931 Simulation information:
    Engine: slim (4.0.1)
    Model id: PiecewiseConstant
    Model description: Piecewise constant size population model over multiple epochs.
    Seed: 1234
    Population: number_samples (sampling_time_generations):
        pop_0: 5 (None)
Contig Description:
    Contig origin: chr1:0-248956
    Contig length: 248956.0
    Contig ploidy: 2
    Mean recombination rate: 1.1523470111585671e-08
    Mean mutation rate: 1.29e-08
    Genetic map: None