Investigate long regendoc output in cache.rst

[The-Compiler]

Hmmm with the documentation moved we are getting cache data from other runs. @RonnyPfannschmidt can we use "regendoc wipe" to fix this?

Originally posted by @nicoddemus in #8814 (comment)

[The-Compiler]

I have no idea on this one - @nicoddemus @RonnyPfannschmidt could one of you take a look at what's wrong there?

[RonnyPfannschmidt]

Slotted for the evening

[RonnyPfannschmidt]

so much for that evening, fetching the details now

[RonnyPfannschmidt]

discovered a bug in bestrelpath

when it goes up to the root, then it will return a relative path that is longer than the rest

[RonnyPfannschmidt]

@bluetech i believe the issue could have gotten introduced while changing our commonpath usage to the stdlib, creating a test+fix now

[RonnyPfannschmidt]

@The-Compiler i made the fixture output better by fixing the bestrelpath for root == commonpath on posix,

additionally im introducing a PYTHON_SITE meta env var for better handling

[RonnyPfannschmidt]

@The-Compiler i discovered that for some reason the python prefix is used as pytest cache dire

[RonnyPfannschmidt]

it seems this first cropped up in 2019 and was back then incorrectly resolved by adding a rm -rf

[The-Compiler]

I don't really follow. @RonnyPfannschmidt what's the state of this? FWIW I don't think this should block a release candidate, but I do think we should fix it for the final.

[RonnyPfannschmidt]

I have not been able to allocate time to go down the rabbit hole, I suspect a path normalising bug, wipe won't help as the pytest cache dir completely miss detected, its absolutely unclear

[nicoddemus]

wipe won't help as the pytest cache dir completely miss detected, its absolutely unclear

You mean deleting .pytest_cache didn't work, or were you unable to do it while running regendoc? Seems just deleting would solve it, otherwise from where would that cache information come from?

[RonnyPfannschmidt]

@nicoddemus the pytest_cache is not in a regendoc tmpdir but in the actual tox virtualenv, i have no idea how/why

when the error first appeared it was simply fixeds in a very broken way, and now with the renamed files its just hitting harder/worse

[nicoddemus]

Ahh I see, thanks (now I recall you mentioning it in that sense elsewhere).