.dockerignore

# Everything the image must not carry, and the reason it matters.
#
# There was no .dockerignore at all, so `docker build` sent the whole working
# tree — including `apps/forum/.env`, a *developer's* file that is gitignored
# and sets QUEUE_DRIVER=postgres. Next loads it during the build, so the
# prerender resolved a Postgres queue while DATA_SOURCE was fixture and the
# build died on `getDb() called while DATA_SOURCE is "fixture"`. The image
# therefore built on a machine with no .env and failed on one with it — and
# baked whatever secrets that file held into any image that did build.
#
# Nothing here is an optimisation. Each line is something that changes what the
# image *is*.

# Local environment. The image is configured by the orchestrator, never by a
# file that happened to be on the machine that built it.
#
# The `**/` matters and is the whole bug: dockerignore patterns are anchored to
# the build context, so a bare `.env` matches only the repo root — and the file
# that broke the build was `apps/forum/.env`, which Next loads and which set
# QUEUE_DRIVER=postgres.
**/.env
**/.env.*
!**/.env.example

# Installed and built artefacts: the image installs and builds its own.
node_modules
**/node_modules
.next
**/.next
**/.next-e2e
**/dist
apps/worker/dist

# forum-web's own materialized app (see docker/Dockerfile's build stage,
# which regenerates it fresh with `pnpm --filter @meith/board-stock build`).
# A developer's own local run of it under boards/stock/.meith would otherwise
# ride into the build context for no reason.
**/.meith

# History and CI: not needed to build, and .git is large.
.git
.github

# Test and tooling output.
coverage
playwright-report
test-results
*.tsbuildinfo

# Editor and OS noise.
.DS_Store
.vscode
.idea