tmp_path symlink oddity

[nicoddemus]

Hi folks,

While investigating #10896, I noticed something which I find odd.

To reiterate how things work: when we use a tmp_path fixture for the first time, we create the basetemp directory using make_numbered_dir, creating a new base temporary directory exclusive for that session:

E:\.tmp\pytest-of-bruno\pytest-905
        ^^^^^^^^^^^^^^^ ^^^^^^^^^^ 
        root dir        basetemp

That root dir is created once for each user, and basetemp is created for each pytest session.

In addition to that pytest-905 directory, it will also create a symlink to it named pytest-current in the same directory (which I'm not sure what the use case is).

E:\.tmp\pytest-of-bruno\pytest-905
E:\.tmp\pytest-of-bruno\pytest-current  <- symlink to pytest-905

So when a test with tmp_path executes, the fixture creates a new temporary directory for that test inside basetemp, again using make_numbered_dir, which also includes that symlink:

E:\.tmp\pytest-of-bruno\pytest-908\test1_7_0
E:\.tmp\pytest-of-bruno\pytest-908\test1_7_current  <- symlink to test1_7_0

I kinda understand the reason for that first symlink to basetemp (pytest-current), but can anyone explain the reason why we have a symlink for each directory created by tmp_path (test1_7_current, above)?

[RonnyPfannschmidt]

It's a make numbered dir convenience going back to debugging pypy tests in pdb

We can probably drop it over printing the paths these days

[nicoddemus]

Thanks @RonnyPfannschmidt

We can probably drop it over printing the paths these days

Drop the top-level symlink (pytest-905 in the example above), the symlinks to the test directories (test_1_7_0), or both?

[RonnyPfannschmidt]

Both