TST/BF: enable overwriting downloads under Windows

[adswa]

This PR aims to fix a test failure that surfaced when a re-download attempts to overwrite an existing file (#5199 (comment)). On Windows systems, an os.rename() will always raise a FileExistsError
if the file exits, while on Unix systems the existing file is replaced silently if the user has permissions, see https://docs.python.org/3/library/os.html#os.rename.
os.replace() appears to be a cross-platform compatible solution, but let's hope that this doesn't break something else - then I could just do it if on_windows.

[adswa]

mental note: remove the known_failure_githubciwin annotation of test_HTTPDownloader_Basic and test_download_url if this fixes the error.

[adswa]

yay, the two tests in question (test_HTTPDownloader_Basic and test_download_url) now pass. I have removed the knownfailure annotation for the GH workflow and will see what happens

[codecov]

Codecov Report

Merging #5201 (3a49df7) into maint (a92ec4c) will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##            maint    #5201      +/-   ##
==========================================
- Coverage   89.90%   89.90%   -0.01%     
==========================================
  Files         294      294              
  Lines       40953    40951       -2     
==========================================
- Hits        36820    36818       -2     
  Misses       4133     4133              

Impacted Files	Coverage Δ
datalad/downloaders/tests/test_http.py	88.59% <ø> (-0.06%)	⬇️
datalad/downloaders/base.py	80.70% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a92ec4c...3a49df7. Read the comment docs.

[kyleam]

Thanks, @adswa. I wasn't aware of os.replace.

yay, the two tests in question (test_HTTPDownloader_Basic and test_download_url) now pass. I have removed the knownfailure annotation for the GH workflow and will see what happens

Hmm, do you mean by looking at the builds for this PR? I think we haven't had a windows build for the github ci since gh-4303.

[adswa]

Hmm, do you mean by looking at the builds for this PR? I think we haven't had a windows build for the github ci since gh-4303.

oh, that's a misconception of mine then, sorry. I thought that this would be the CrippledFS workflow. Shall I drop that commit (4a09de1)?

[kyleam]

I thought that this would be the CrippledFS workflow.

That's ubuntu with $TMPDIR pointing to a vfat fs (with the idea that that will reveal things like adjusted branch issues that aren't windows-specific).

Shall I drop that commit (4a09de1)?

Given it fixed these test failures on your end, I'd say it's fine to keep. If at some point a github ci windows build is re-enabled, we'll need assess where things stand anyway, so the worst case is that these two tests need to be marked up again.

[kyleam]

Any objections to moving this to maint?

[adswa]

Any objections to moving this to maint?

Not at all. I have to admit, I too rarely think of this in advance. I can rebase shortly, will just finish a commit on another branch

[kyleam]

@adswa I'm guessing the build failures are from a push followed by changing the base on github rather than the other way around. Triggering a new build (with a noop rewrite or a rebase to the latest maint) would hopefully sort it out. Sorry for the hassle.

[adswa]

I have rebased to the latest maint