Mirrors/etherpad-lite
1
0
Fork 0
mirror of https://github.com/ether/etherpad-lite.git synced 2026-07-17 16:47:05 +00:00
Code Issues Projects Releases Wiki Activity

fix(docker): bypass pnpm at runtime to avoid spurious deps-status reinstall (#7718) (#7727)

Browse source 
* chore: ignore /.worktrees/ for local worktree workflows

* fix(docker): bypass pnpm at runtime to avoid spurious deps-status reinstall (#7718)

pnpm 11's runDepsStatusCheck runs before every `pnpm run …` and decides
node_modules is out of sync on container first start under the named-
volume layout used by docker-compose (mounting src/plugin_packages). It
then spawns `pnpm install --production`, which either prompts to wipe
node_modules (tty: true) or aborts with
ERR_PNPM_ABORTED_REMOVE_MODULES_DIR_NO_TTY (no tty).

Reproduced by kimllee in ether/etherpad#7718 with the official
etherpad/etherpad:latest image on arm64.

Run node directly in CMD instead of going through `pnpm run prod`.
The image's node_modules was already verified during build, so the
runtime check adds no value. Wrapping in `sh -c 'cd src && exec node …'`
keeps WORKDIR consistent for `docker exec` users while making node PID 1
so it receives SIGTERM directly and shuts down cleanly.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* ci(docker): regression test for #7718 — boot with named volume on plugin_packages

Reproduces the production docker-compose layout from #7718: a named
volume on src/plugin_packages and no allocated TTY. Under the previous
`CMD ["pnpm", "run", "prod"]`, pnpm 11's runDepsStatusCheck spuriously
flagged node_modules out of sync at boot, spawned `pnpm install
--production`, and aborted with ERR_PNPM_ABORTED_REMOVE_MODULES_DIR_NO_TTY
before the HTTP server came up.

If the Dockerfile CMD is ever reverted to invoke pnpm at runtime, this
step times out waiting for the health endpoint and fails CI.

Addresses Qodo review feedback on #7727.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
John McLear 2026-05-11 19:13:43 +01:00 • committed by GitHub
parent 3113c5096f
commit c7100e0ba4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 51 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

36
.github/workflows/docker.yml vendored
View file

@ -79,7 +79,43 @@ jobs:
esac
done
(cd src && pnpm run test-container)
docker rm -f test
git clean -dxf .
-
# Regression test for #7718. Reproduces the production
# docker-compose layout reported in that issue: a named volume
# mounted on src/plugin_packages with no TTY allocated. Under
# the previous `CMD ["pnpm", "run", "prod"]`, pnpm 11's
# runDepsStatusCheck spuriously decided node_modules was out of
# sync at boot and tried to wipe + reinstall, aborting with
# ERR_PNPM_ABORTED_REMOVE_MODULES_DIR_NO_TTY before the HTTP
# server ever came up. If the CMD is ever reverted to go
# through `pnpm run`, this step will time out waiting for the
# health endpoint and fail.
name: Regression — boot with named volume on plugin_packages (#7718)
working-directory: etherpad
run: |
docker volume create ep7718-plugins
docker run --rm -d -p 9001:9001 \
-v ep7718-plugins:/opt/etherpad-lite/src/plugin_packages \
--name test-7718 ${{ env.TEST_TAG }}
docker logs -f test-7718 &
ok=0
for i in $(seq 1 60); do
status=$(docker container inspect -f '{{.State.Health.Status}}' test-7718 2>/dev/null) || {
echo "container exited prematurely — likely #7718 regression"
docker logs test-7718 || true
break
}
case ${status} in
healthy) ok=1; echo "container healthy after $((i*2))s"; break;;
starting) sleep 2;;
*) echo "unexpected status: ${status}"; docker logs test-7718; break;;
esac
done
docker rm -f test-7718 >/dev/null 2>&1 || true
docker volume rm ep7718-plugins >/dev/null 2>&1 || true
[ "$ok" = "1" ] || exit 1
build-test-local-plugin:
# Regression coverage for #7687: the Docker image's