fix: ignoring symlinks

[gabyx]

@brianmcgee : I tried the following:

• Ignore symlinks on the Git reader.
• Adding a symlink to the examples, to test if it gets ignored everywhere this examples are used.
• Add resolving symlinks on the Run functions to pass the resolved paths to the readers, such that the readers Filesystem or Git can ignore symlinks always. treefmt symlink.yaml or treefmt /a-symlink/b/c/d.yaml should be allowed.

See some comments I need help on which is unclear to me.

Fixes: #495

[gabyx]

@brianmcgee : I have question to the test here:

Why do we even test to run in the .git folder? You will never want that IMO, or is this really a feature?
Shouldn't treefmt ignore any .git directory (got from git rev-parse --git-common-dir in the tree root)

treefmt/cmd/root_test.go

Line 1401 in bdd8687

// walk with filesystem instead of with git

[brianmcgee]

Why do we even test to run in the .git folder? You will never want that IMO, or is this really a feature? Shouldn't treefmt ignore any .git directory (got from git rev-parse --git-common-dir in the tree root)

In that particular test, it's checking that if we momentarily switch to the filesystem walker, the .git directory gets picked up and traversed. It's not that we want to traverse the .git directory per se, but that it should traverse any directory when the filesystem walker is enabled.

[gabyx]

Why do we even test to run in the .git folder? You will never want that IMO, or is this really a feature? Shouldn't treefmt ignore any .git directory (got from git rev-parse --git-common-dir in the tree root)

In that particular test, it's checking that if we momentarily switch to the filesystem walker, the .git directory gets picked up and traversed. It's not that we want to traverse the .git directory per se, but that it should traverse any directory when the filesystem walker is enabled.

Ah ok.

I need to still refactor the symlink resolving code in the Run function, its not good yet. I think the only correct way is to us EvalSymlinks/ or only os.Stat and check for errors and make the test pass. As this is only done on path inputs it should not degrade performance. ki

[gabyx]

@brianmcgee: The thins is done. Happy for a review what you think about the behavior. Hope its ok like that.

[brianmcgee]

Thanks for this, I'll try and review it in the next few days.

[brianmcgee]

@gabyx please rebase on latest main

[gabyx]

@brianmcgee: Did the changes and rebased.

[brianmcgee]

Thanks 🙏 I'll give it another review in the next day or so