[stevenchen-db]

Changes

databricks air run no longer needs git on PATH to package a code_source directory that isn't in a git repository.

• Outside a repository, no git: the directory is packaged with libs/git's ignore rules (root and nested .gitignore, info/exclude, core.excludesFile). Symlinks are listed, not followed; a nested repository is one entry, as with git.
• Inside a repository, or code_source.git set, no git: fails before validation, packaging, or any API call, with an actionable error.
• git on PATH: the existing git-based listing is unchanged.

 $ PATH=<no git> databricks air run -f workload.yaml --dry-run    # code_source.root_path: . (not a repo)
-Error: failed to initialize temporary git metadata: exec: "git": executable file not found in $PATH:
+Dry run: code_source snapshot would include 2 file(s).

 $ PATH=<no git> databricks air run -f workload.yaml --dry-run    # code_source inside a git repo
-Error: failed to initialize temporary git metadata: exec: "git": executable file not found in $PATH:
+Error: code_source at /path/to/repo is in a git repository, so packaging it needs git on PATH; install git and retry

Known differences of the git-free listing (only used without git; pinned by tests): a root rule wins over a nested negation (*.log at the root excludes nested/keep.log even with !keep.log in nested/.gitignore), and include_paths globs use path.Match, where * doesn't cross /.

Why

Hosts without git (containers, CI images, agent sandboxes) couldn't use code_source at all, though git is only needed for its ignore rules outside a repository. The quickstart lists git as a requirement; a docs follow-up can narrow that to code inside a git repository.

No Jira ticket.

Tests

• Unit: the git-free listing matches git ls-files for whole trees, directory and file include_paths, nested .gitignore, negation, ignored dirs, dotfiles and symlinks; nested repo as one entry; the two known differences; the early error for in-repo and pinned code (with an injected git lookup, also asserting no lookup without code_source).
• Real workspace, no git on PATH: a code_source run succeeded and printed its output (run 892749685192587); the in-repo case failed with 0 API calls; a workload without code_source is unchanged. With git on PATH, the dry-run file count and request list match main.
• TestAccept/air/ 20/20, lint clean, ./task checks passes. TestLogsCommandRepeatsRunLinkAfterStreaming fails the same way on main and is unrelated.

This pull request and its description were written by Isaac.

[eng-dev-ecosystem-bot]

✅ Integration test report

Commit: f74898f

Run: 43534193924

	Env	💚​RECOVERED	✅​pass	🙈​skip	Time
💚​	aws linux-2core-8gb	1	270	15	6:15
💚​	aws-windows-latest-4core-16gb	1	272	13	3:38
💚​	azure linux-2core-8gb	1	269	15	6:22
💚​	azure-windows-latest-4core-16gb	1	271	13	3:50
💚​	gcp linux-2core-8gb	1	270	15	7:56
💚​	gcp-windows-latest-4core-16gb	1	272	13	5:11

	Test Name	aws linux-2core-8gb	aws-windows-latest-4core-16gb	azure linux-2core-8gb	azure-windows-latest-4core-16gb	gcp linux-2core-8gb	gcp-windows-latest-4core-16gb
💚​	TestAccept	💚​R	💚​R	💚​R	💚​R	💚​R	💚​R

Top 4 slowest tests (at least 2 minutes):

duration	env	testname
3:13	gcp-windows-latest-4core-16gb	TestAccept
3:13	aws-windows-latest-4core-16gb	TestAccept
2:44	azure-windows-latest-4core-16gb	TestAccept
2:13	gcp linux-2core-8gb	TestSyncIncrementalFileSync