gitignore doesn't work if atom is opened within subdirectory of git repo

[aaronjensen]

I don't know if this is a fuzzy-finder or a core problem, but if I'm in a subdirectory of a git repo when I open atom the gitignore is... ignored.

[maaarghk]

See issue #111 - related?

[aaronjensen]

hmm, related in that the two issues kinda seem at odds to each other. My problem is that the parent dir gitignore isn't used.

Example:

.gitignore:

node_modules

folder structure:

.gitignore
client
-- node_modules
server
-- node_modules

If I open atom client then node_modules is not ignored. If I open atom . then node_modules is ignored.

[manishtpatel]

+1, even adding .gitignore at client folder level does not help either.

[opensourceportfolio]

same issue here

[andykenward]

Having the same issue working in a sub directory of a repo that has its own .gitignore.
Changing to root of the repo solves the issue for me.

Atom 1.0.7

[nielsAD]

So this appears to be expected behaviour, according to the tests. I'm not sure if it's desired behaviour, though, as tree-view appears to flag the files as ignored.

[aaronjensen]

I definitely do not desire it :)

[nielsAD]

It looks like tree-view's "Hide VCS Ignored Files" setting has a similar behaviour; items are only hidden when you are actually working in the root of the VCS. Ignored files are still greyed out, though, when you are working in a subdirectory.

I'm not sure what the best solution would be... Maybe an extra setting? Or maybe the sorting should take ignored files into account and place them lower on the list? #133 is a PR that just disables the root folder check, so that would result in some inconsistencies between tree-view and fuzzy-finder.

[maaarghk]

Surely they would ideally both act in the same way which the relevant VCS works? An extra setting would be strange, it'd be an option to be inconsistent and therefore not useful. Inconsistencies between tree-view and fuzzy-finder are okay if one is wrong. Just means an issue needs to be raised on the incorrect module (this one, in my view)

[nielsAD]

The best solution I can think of is that we let it depend on the working directory. If that one is ignored by the VCS, we do not use the VCS ignores and just display everything. Otherwise, we inherit the VCS ignores.

[maaarghk]

Sort of - a working directory which is itself .gitignored, but has its own .git folder (i.e. git submodule), should not process the parent .gitignore. It should use the .gitignore of the working directory (but not ignore it.)