air: reject code_source include_paths entries that match no files

.nextchanges/air/7639.md

@@ -0,0 +1 @@
+* `air run` now rejects a `code_source.include_paths` entry that matches no files instead of uploading a snapshot without it. ([#7639](https://github.com/databricks/cli/pull/7639))

cmd/air/snapshot_package.go

@@ -8,6 +8,7 @@ import (
 	"io"
 	"os"
 	"os/exec"
+	"path"
 	"path/filepath"
 	"strings"
 
@@ -193,9 +194,39 @@ func snapshotFiles(ctx context.Context, repoPath string, includePaths []string,
 	if len(files) == 0 {
 		return nil, emptySnapshotError(ctx, repoPath, includePaths, isGitRepo)
 	}
+	if unmatched := unmatchedIncludePaths(includePaths, files); len(unmatched) > 0 {
+		return nil, fmt.Errorf("code_source include_paths select no files under %s: %s. Check that each path exists under root_path and is not ignored by git (`git check-ignore -v <path>`)", filepath.ToSlash(repoPath), strings.Join(unmatched, ", "))
+	}
 	return files, nil
 }
 
+// unmatchedIncludePaths returns the include_paths entries that select none of files.
+// git ls-files ignores a pathspec that matches nothing, so a mistyped entry would be
+// left out of the snapshot and the run would fail on the node after compute was
+// provisioned. Entries with glob characters are skipped because git matches them as
+// patterns rather than as path prefixes.
+func unmatchedIncludePaths(includePaths []string, files []snapshotFile) []string {
+	var unmatched []string
+	for _, p := range includePaths {
+		if strings.ContainsAny(p, "*?[") {
+			continue
+		}
+		prefix := path.Clean(filepath.ToSlash(p))
+		matched := prefix == "."
+		for _, file := range files {
+			rel := filepath.ToSlash(file.rel)
+			if rel == prefix || strings.HasPrefix(rel, prefix+"/") {
+				matched = true
+				break
+			}
+		}
+		if !matched {
+			unmatched = append(unmatched, p)
+		}
+	}
+	return unmatched
+}
+
 // emptySnapshotError explains why no file was selected for the snapshot. An empty
 // tarball is accepted by the upload but fails on the node, after compute was
 // provisioned, so it is rejected here with the most likely cause.

cmd/air/snapshot_package_test.go

@@ -324,3 +324,42 @@ func TestSnapshotFiles_RejectsIncludePathsMatchingNothing(t *testing.T) {
 	assert.Contains(t, err.Error(), "contains no files")
 	assert.Contains(t, err.Error(), "include_paths")
 }
+
+func TestSnapshotFiles_RejectsIncludePathMatchingNothingAlongsideOthers(t *testing.T) {
+	repo := t.TempDir()
+	writeRepoFile(t, repo, "ray_bootstrap.sh", "#!/bin/bash\n")
+	writeRepoFile(t, repo, "my-job/main.py", "print(1)\n")
+
+	_, err := snapshotFiles(t.Context(), repo, []string{"ray_bootstrap.sh", "my_job"}, false)
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), "include_paths select no files")
+	assert.Contains(t, err.Error(), ": my_job.")
+	assert.NotContains(t, err.Error(), "ray_bootstrap.sh")
+}
+
+func TestSnapshotFiles_RejectsIncludePathWhoseFilesAreIgnored(t *testing.T) {
+	repo := newTestRepo(t)
+	writeRepoFile(t, repo, ".gitignore", "build/\n")
+	writeRepoFile(t, repo, "main.py", "print(1)\n")
+	writeRepoFile(t, repo, "build/out.bin", "x")
+
+	_, err := snapshotFiles(t.Context(), repo, []string{"main.py", "build"}, true)
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), ": build.")
+	assert.Contains(t, err.Error(), "check-ignore -v")
+}
+
+func TestSnapshotFiles_AcceptsIncludePathForms(t *testing.T) {
+	repo := t.TempDir()
+	writeRepoFile(t, repo, "src/model.py", "print()")
+	writeRepoFile(t, repo, "README.md", "readme")
+	writeRepoFile(t, repo, "notes.txt", "notes")
+
+	files, err := snapshotFiles(t.Context(), repo, []string{"./src/", "README.md", "*.txt"}, false)
+	require.NoError(t, err)
+	var rels []string
+	for _, file := range files {
+		rels = append(rels, filepath.ToSlash(file.rel))
+	}
+	assert.ElementsMatch(t, []string{"src/model.py", "README.md", "notes.txt"}, rels)
+}