Asset bundles sync don't follow .gitignore pattern syntax for root folders

[tom-rb]

Describe the issue

The sync: include: configuration of asset bundles don't respect leading slash / as .gitignore does:

From git docs (my highlight):

• The slash "/" is used as the directory separator. Separators may occur at the beginning, middle or end of the .gitignore search pattern.
• If there is a separator at the beginning or middle (or both) of the pattern, then the pattern is relative to the directory level of the particular .gitignore file itself. Otherwise the pattern may also match at any level below the .gitignore level.

I wanted to force an include a folder present in the root of my repo, but all folders named the same (inside other .gitignored folders) were included in the sync.

Configuration

File tree

.
├── ignored
│   └── my_folder
└── my_folder

databricks.yml

sync:
  include:
    - /my_folder/

.gitignore

/ignored
/my_folder

Result: both my_folder gets uploaded to Databricks workspace

Steps to reproduce the behavior

1. Prepare files as above
2. Run databricks bundle deploy
3. See error

Expected Behavior

Only the my_folder in the root of the repo should be synced.

Actual Behavior

All folders named my_folder are synced.

OS and CLI version

Linux 6.6.87.2-microsoft-standard-WSL2
Databricks CLI v0.260.0

Is this a regression?

I haven't tested

Debug Logs

–

[github-actions]

This issue has not received a response in a while. If you want to keep this issue open, please leave a comment below and auto-close will be canceled.

[tom-rb]

I find this still relevant, the bot wants me to comment.

[TangoEnSkai]

I could not reproduce this against main. Anchored sync.include patterns do behave like gitignore today.

sync.include reaches fileset.NewGlobSet, which compiles the patterns with go-gitignore, so a leading slash anchors to the fileset root. Running your exact tree (my_folder/ and ignored/my_folder/) through NewGlobSet:

include="/my_folder/"  -> [my_folder/a.txt]
include="my_folder/"   -> [ignored/my_folder/b.txt my_folder/a.txt]

So /my_folder/ selects only the root directory, and dropping the slash is what makes it match at any depth — which is the behaviour you expected.

Worth noting the matching code has not meaningfully changed since v0.260.0, the version in the report: NewGlobSet did the same path.Clean + CompileIgnoreLines then. The only later edit was an early return for the empty-pattern case. So this probably was not a matching bug at the time either.

Two things that could explain what you saw, if you still have the repro:

1. Files also arrive from the git-tracked set, not just include. FileList.Files returns (git ∪ include) − exclude. If ignored/ was not actually ignored by git at deploy time — for example the entry was added after those files were already tracked — ignored/my_folder would be synced regardless of what sync.include says. git check-ignore -v ignored/my_folder would confirm.
2. sync.paths selects which subtrees are listed at all, independently of include.

Could you check whether the files were git-tracked in your repro? If they were, this is working as intended and the issue can be closed; if not, a minimal repro would help a lot, since the pattern layer looks correct.

Separately, I noticed nothing in the test suite pinned this anchoring behaviour, so I opened #6345 to add a regression test for it — trimming the leading slash inside NewGlobSet currently breaks no existing test.