Faked os.makedirs differs from the built-in os.makedirs and mkdir -p

[vector-of-bool]

Describe the bug

This issue comes with some bonus cursed knowledge.

os.makedirs and mkdir -p both do a "weird" thing when they create directory trees:

$ mkdir -p base/foo/../bar
# ls base
bar/  foo/

That is:

1. It creates base/ if absent, then enters base/
2. Create foo/ if absent, then enter foo/
3. It creates ../ if absent (it is never absent on "normal" filesystems), then enters ../ (sending it back into base/)
4. Then creates bar/

(os.makedirs does the same thing.)

pyfakefs's os.makedirs normalizes the path before it does the above process, meaning the foo and .. parts will be chopped out of the path, and the foo/ directory will not be created.

How To Reproduce

def weird_makedirs(base: Path):
    shutil.rmtree(base, ignore_errors=True)
    b = base / "foo/../bar"
    os.makedirs(b)
    assert (base / "bar").is_dir()
    assert (base / "foo").is_dir()
    assert not (base / "foo/bar").is_dir()

def test_weird_makedirs_realfs():
    weird_makedirs(Path("testdir-tmp"))

def test_weird_makedirs_fakefs(fs):
    weird_makedirs(Path("testdir-tmp"))

Your environment
Please run the following in the environment where the problem happened and
paste the output.

$ python -c "import platform; print(platform.platform())"
Linux-6.7.6-arch1-1-x86_64-with-glibc2.39
$ python -c "import sys; print('Python', sys.version)"
Python 3.11.7 (main, Jan 29 2024, 16:03:57) [GCC 13.2.1 20230801]
$ python -c "from pyfakefs import __version__; print('pyfakefs', __version__)"
pyfakefs 5.4.dev0
$ python -c "import pytest; print('pytest', pytest.__version__)"
pytest 7.2.0

[mrbean-bremen]

I tried to fix this yesterday, and it worked under Linux, but not under Windows (that is, my test didn't, which had more directories as base, yours did). I did not understand yet the real behavior under Windows - it is somewhat more permissive than under Linux, but not always. I will probably not be able to fix this in the next couple of weeks, as I have near to no free time now, and will be away afterwards. I guess I will find more of your issues when I get back :)

[mrbean-bremen]

I actually had made a testing error, so this should be good now.

[mrbean-bremen]

@vector-of-bool - I'm back, did you have the the time to test some more? I'm inclined to make a new release if nothing else comes up.

[vector-of-bool]

I've been bouncing between (side) projects and haven't had a chance to test this one yet. I'll try to get back on this one sooner than later.

[mrbean-bremen]

I'll probably make a release at the weekend anyway, if you don't get the time to test it. There have been some fixes accumulated that are worth publishing, and we can always make patch releases if something else comes up.
Update: done.