Background
The repository has a 326 MB docs/ directory (JDK 25 API/spec HTML docs used as the RAG corpus) that is entirely untracked and not listed in .gitignore. The workshop branch historically ignored docs/; main does not. A single git add -A would stage 326 MB of generated documentation.
Impact
- One careless command commits ~326 MB of JDK HTML docs, permanently bloating the repository and its history.
- Clone times and CI checkout times degrade for every contributor.
- Large files can hit GitHub's per-file size limits and cause push failures.
How to reproduce
- From the repo root, run
git add -A.
- Observe
docs/ being staged (~326 MB).
Where the fix should land
Add docs/ to .gitignore with a comment explaining it is the local JDK-docs corpus provisioned on the VM / mounted via Docker. Optionally commit docs/.gitkeep or docs/README.md documenting how to obtain the corpus, so the directory structure is documented without shipping the content.
Note: docker-compose.yml already mounts ./docs:/app/docs:ro — the ignore does not affect that volume mount.
Files touched
.gitignore
docs/.gitkeep or docs/README.md (optional)
Acceptance criteria
Background
The repository has a 326 MB
docs/directory (JDK 25 API/spec HTML docs used as the RAG corpus) that is entirely untracked and not listed in.gitignore. The workshop branch historically ignoreddocs/;maindoes not. A singlegit add -Awould stage 326 MB of generated documentation.Impact
How to reproduce
git add -A.docs/being staged (~326 MB).Where the fix should land
Add
docs/to.gitignorewith a comment explaining it is the local JDK-docs corpus provisioned on the VM / mounted via Docker. Optionally commitdocs/.gitkeepordocs/README.mddocumenting how to obtain the corpus, so the directory structure is documented without shipping the content.Note:
docker-compose.ymlalready mounts./docs:/app/docs:ro— the ignore does not affect that volume mount.Files touched
.gitignoredocs/.gitkeepordocs/README.md(optional)Acceptance criteria
git check-ignore docs/reports the directory as ignored.git add -Ano longer stagesdocs/.docs/README.mdexplains where the corpus comes from.