Can't write first entry in directory journal when directory doesn't exist about jrnl HOT 3 CLOSED

Looks like this is Windows-only, since a related CI test is failing only on Windows.

I also couldn't replicate this locally on WSL/Ubuntu. jrnl --diagnostic output:

jrnl: v4.1
Python: 3.11.6 (main, Oct  2 2023, 13:45:54) [GCC 11.4.0]
OS: Linux 5.15.146.1-microsoft-standard-WSL2

from jrnl.

I'm mystified by this bug. os.makedirs is supposed to create all directories in the path that don't exist, so why is Windows kicking out an error about it not existing? Furthermore, we have a passing BDD test in file_storage.feature for exactly this problem.

I was able to reproduce this locally for a little while. I also reproduced it in the interactive Python 3.12 shell with os.makedirs. Then I did something--I'm not sure what--and now it just works. What changed? Maybe a permissions issue on one of those temp subfolders? But why would that have existed in the first place? And why does this appear in GitHub Actions, but only for a new PR?

I searched around for others with issue, but most of the resolutions I see also don't seem to make much sense, and I think they're not actually solving the problem so much as drawing post hoc conclusions. I don't see anyone writing code to fix this problem.

Whatever the case, that brief test that reproduced this problem in interactive Python showed that it's a problem with Windows (or Python on Windows), rather than a problem with jrnl. Python on Windows is failing to deliver what the os.makedirs documentation promises, or, at the least, it's failing to accurately report the reason it can't meet that promise, so I'm closing this as wont-fix.

If you're a jrnl user and you have this problem, please let us know here and maybe we can narrow it down some more, but without reports of anyone else running into this, I don't think it's worth the effort to keep investigating.

from jrnl.

I've found the root cause for this and it's expected behaviour (though unintuitive behaviour).

From the docs, Jrnl detects that a journal is a Folder journal if

• a folder with the name of the journal before running jrnl.
• the journal path in the config ends with the path separator

In the empty_folder config, the journal path is specified without a path separator at the end. To fix the test failures we could add the path separator to the path in the config but then two configs would be needed (one for Posix, one for

A proper fix would be to use pathlib instead of os.path since that translates Posix paths into windows paths when required.

The fact that the tests work on Linux is the actual bug. Running

jrnl --cf tests/data/configs/empty_folder.yaml --debug "this is my jrnl entry"

shows that, on Linux, it creates a FileJournal and happily chugs along.

from jrnl.