Bug: Git-ignored directory paths are treated as unanchored patterns, excluding unrelated source files #2849

Environment

• CodeGraph: 1.6.2
• OS: Linux
• Node.js: 24.14.1
• Repository: Java multi-module Maven monorepo

Problem

CodeGraph incorrectly excludes thousands of tracked Java files when certain valid .gitignore patterns are present.

For example:

*jib*
**/opt/datadog

With these patterns, CodeGraph indexes zero Java files. Removing them allows 3,368 Java files to be indexed.

Root cause

In lib/dist/extraction/index.js, buildScopeIgnore() adds Git-reported ignored directory paths directly to the ignore matcher:

for (const dir of listGitIgnoredDirectories(rootDir)) {
    rootMatcher.add(dir);
}

Git legitimately reports the root-relative ignored directory src/.

However, the ignore library interprets src/ as matching directories with that name at any depth.

Consequently, CodeGraph incorrectly excludes paths such as:

services/backend-for-frontend/src/main/java/...

even though Git tracks these files and does not ignore them.

Proposed fix

Anchor Git-reported directory paths to the project root:

for (const dir of listGitIgnoredDirectories(rootDir)) {
-    rootMatcher.add(dir);
+    rootMatcher.add('/' + dir.replace(/^\/+/, ''));
}

Verification

Before patch:

• 257 indexed files
• 0 Java files

After patch:

• 3,368 Java files indexed
• Original .gitignore restored
• Root src/ remains ignored
• Nested service Java sources are correctly included

The fix was verified against CodeGraph 1.6.2 using buildScopeIgnore() and a full index rebuild.

Suggested regression test

Create a Git repository containing:

src/main/jib/opt/datadog/dd-java-agent.jar
services/example/src/main/java/Example.java

Add a .gitignore rule that ignores the JAR or its directory.

Assert that services/example/src/main/java/Example.java remains discoverable and indexable.