RF: Cleanup Makefile to bring parity with CI, remove tox + nose references

[jacobdr]

Closes #1104

• Align on venv as virtual environment convention
• Update makefile test target to re-use same script used in CI to orchestrate testing
• Remove refererences to tox and nose
• Fix dev-requirements.txt and doc-requirements.txt to pin to versions that are non-breaking

[pep8speaks]

Hello @jacobdr, Thank you for updating!

Cheers! There are no style issues detected in this Pull Request. 🍻 To test for issues locally, pip install flake8 and then run flake8 nibabel.

Comment last updated at 2022-05-07 19:05:52 UTC

[codecov]

Codecov Report

Merging #1105 (ad2cc68) into master (7cfaebf) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #1105   +/-   ##
=======================================
  Coverage   92.29%   92.29%           
=======================================
  Files         100      100           
  Lines       12247    12247           
  Branches     2393     2393           
=======================================
  Hits        11303    11303           
  Misses        621      621           
  Partials      323      323           

---

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 7cfaebf...ad2cc68. Read the comment docs.

[jacobdr]

I wasn't sure if this target is still being used anywhere -- so I kind of hacked it to install the dev dependencies to get it to pass on my local

[jacobdr]

I needed to pin these to get version compatibility to work. In the future might be better to use a tool like poetry so that we can generate a lockfile and not have to deal with this sort of not-fun dependency pinning

[jacobdr]

I needed this to get my local to pass but it seems like CI might not like it.... I'll try reverting to see if if this is what is breaking the pre-release CI jobs

[jacobdr]

Need a follow-up PR to clean up the remaining issues. I added these as ignores so that local flake8 passes. These were being covered up by the configuration of pep8speaks only looking at the diff

[jacobdr]

This change was made so that we are consistent in local dev and remote dev with the virtualenv

[jacobdr]

I removed this as I did not see it being used anywhere. I can revert this if a maintainer feels it should be brought back

[jacobdr]

@effigies all the pre-release tests failed for the same reason -- a string representation in a doctest for netcdf_file. I don't immediately see how my changes would have introduced this. Looking at this now, but any thoughts as to root cause?

[jacobdr]

Trying to narrow differences... since mmap is coming from the standard lib I suspected a change in Python versions and how repr(mmap_obj) or str(mmap_obj) is displayed... Its possible, though still smells to me so needs a sanity check.

As evidence, though, comparing against a prior (chose #1096 as the most recent unmerged green build I saw....)

Python version comparison

• 3.10.4 this MR vs 3.10.2 prior success
• 3.9.12 this MR vs 3.9.10 prior success
• the scheduled pre-release cron job against master is failing for the same reason.

Final note: the only pre-release job that passes is the one that does not have the PRE_PIP_FLAGS environment variable set....

[effigies]

I would rather put make style in check.sh than call check.sh from here. And likewise with other tests.

Suggested change

	.PHONY: test-style
	test-style:
	export CHECK_TYPE=style; ./tools/ci/check.sh
	.PHONY: style
	style:
	flake8