Never sync code files to blob storage

syncmedia.go

@@ -48,6 +48,10 @@ const (
 
 // mediaExtensions is the set of file extensions treated as "media" (i.e.
 // non-code) that will be synced to blob storage. Keep in lowercase.
+//
+// This is an allowlist: files whose extensions are not listed here are never
+// uploaded. See codeExtensions for an explicit denylist of source-code
+// extensions that must never leak into storage even if this allowlist grows.
 var mediaExtensions = map[string]bool{
 	".png":  true,
 	".jpg":  true,
@@ -71,6 +75,27 @@ var mediaExtensions = map[string]bool{
 	".zip":  true,
 }
 
+// codeExtensions is an explicit denylist of source-code file extensions that
+// must NEVER be uploaded to blob storage. Code lives in git; only rendered
+// media belongs in the container. This check runs before the mediaExtensions
+// allowlist so accidental additions to that allowlist can't slip code through.
+// Keep in lowercase.
+var codeExtensions = map[string]bool{
+	".md":   true,
+	".css":  true,
+	".html": true,
+	".go":   true,
+	".js":   true,
+	".ts":   true,
+	".json": true,
+	".yaml": true,
+	".yml":  true,
+	".toml": true,
+	".mod":  true,
+	".sum":  true,
+	".sh":   true,
+}
+
 // syncMediaPrefixes are the repo-relative directory roots we sync from.
 var syncMediaPrefixes = []string{
 	"content/articles",
@@ -258,7 +283,15 @@ func watchAndSync(
 }
 
 func isMediaPath(p string) bool {
-	return mediaExtensions[strings.ToLower(filepath.Ext(p))]
+	ext := strings.ToLower(filepath.Ext(p))
+	if codeExtensions[ext] {
+		return false
+	}
+	return mediaExtensions[ext]
+}
+
+func isCodePath(p string) bool {
+	return codeExtensions[strings.ToLower(filepath.Ext(p))]
 }
 
 // isLikelyDirEvent returns true if the event has no file extension (usually a
@@ -293,6 +326,9 @@ func collectLocalMedia(log *logrus.Logger) (map[string]*localMediaFile, error) {
 				return nil
 			}
 			ext := strings.ToLower(filepath.Ext(p))
+			if codeExtensions[ext] {
+				return nil
+			}
 			if !mediaExtensions[ext] {
 				return nil
 			}

syncmedia_test.go

@@ -0,0 +1,42 @@
+package main
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestIsMediaPathAndIsCodePath(t *testing.T) {
+	mediaCases := []string{
+		"content/articles/foo/bar.png",
+		"content/images/logo.JPG",
+		"content/pages/about/pic.jpeg",
+		"content/articles/x/movie.mp4",
+		"content/articles/x/doc.pdf",
+	}
+	for _, p := range mediaCases {
+		require.Truef(t, isMediaPath(p), "expected media: %s", p)
+		require.Falsef(t, isCodePath(p), "expected not code: %s", p)
+	}
+
+	codeCases := []string{
+		"content/articles/foo/foo.md",
+		"content/stylesheets/tailwind.css",
+		"web/html/_nav.tmpl.html",
+		"build.go",
+		"package.json",
+		"go.mod",
+	}
+	for _, p := range codeCases {
+		require.Truef(t, isCodePath(p), "expected code: %s", p)
+		require.Falsef(t, isMediaPath(p), "expected not media: %s", p)
+	}
+
+	// Belt-and-suspenders: even if a code extension were mistakenly added to
+	// mediaExtensions, isMediaPath must still return false because
+	// codeExtensions is consulted first.
+	mediaExtensions[".md"] = true
+	t.Cleanup(func() { delete(mediaExtensions, ".md") })
+	require.False(t, isMediaPath("content/articles/foo/foo.md"),
+		"code denylist must win over media allowlist")
+}