* add: new inputs and method params
* feat: add scoped registry to manifest
* feat: setup test job
* fix(workflow): revert change from license to serial
* feat: support private scoped registries
* fix: multiple scopes
- Updated activation strategy to align with builder strategy for more stable activation with retry logic
- Added `containerRegistryImageVersion` and `containerRegistryRepository` for improved custom image support
- Added `runAsHostUser` to fix self-hosted runner issues
- Scripts folder cleanup
- Added error/warning annotations for activation issues
- Improved serial masking
- Image environment factory addition and docker refactoring
Cancelled or timeouted workflow would keep the docker container running.
Closesgame-ci/unity-test-runner#197
This has two parts:
Part one. The entrypoints.
`runs.post`: GitHub Action metadata allow running something after the
action (regardless of a failure, crash, timeout, ...).
https://docs.github.com/en/actions/creating-actions/metadata-syntax-for-github-actions#runspost
However, it needs to be a `.js` file and it can't be configured to pass
any arguments.
There already was `index.js` used as the main entrypoint.
The build process of this file uses typescript compiler and ncc to pack
all dependencies into one .js file. And ncc has no way of generating
multiple files in one go, so the only solution would be to run ncc twice
and generate two independent files.
That would be quite unfortunate, wasting time and storage. So I rather
came up with a new entrypoint that symlinked from two locations.
And this new entrypoint understands how it was executed, so it can run
the correct behaviour. This makes it easy to add `runs.pre` if needed.
This new entrypoint is in `index.ts`. The original `index.ts` is now in
`main.ts`.
Part two.
The signals. I've tried:
* try/catch/finally around the `await Docker.run`. Catch and finally are
not executed when process receives SIGINT. See the discussion in: https://github.com/nodejs/node/discussions/29480
* New AbortController and AbortSignal. Great concept, but the
action.exec does not support it. So it can't be aborted.
* Doing cleanup on `process.on('exit')`. Unfortunately you can't really
do async stuff from there, so can't really run the docker rm command
to delete the container.
* Using `process.on('SIGINT')`. For some reason that wasn't really
executing for me. I'd not put my hand in fire for this, but I assume
because it was in the signal handler it does something special, or would
heed to be scheduled for later with `setTimeout(0)`.
Evaluating all these I came to a conclusion that it is fragile and just
relying on a `runs.post` is much better and safer.
`
* Refactor to typescript
* Fix platform and workflow
* Fix workflow and platform
* Update husky and image-tag
* Use 'yarn lint-staged' instead of 'npx lint-staged'
* using SSH_AUTH_SOCK (ssh agent forwarding) to pull upm private repos
* sshAgent as input parameter
* yarn run prettier --write "src/**/*.{js,ts}"
* yarn run lint --fix && yarn build
* reverted results-meta.ts (changed because ran prettier --check "src/**/*.js" without && eslint src)
* removed RUN apt-get update && apt-get install -y openssh-client. This change needs to be done upstream. See game-ci/docker#117
nic van dessel
Andrew Kahr
David Finol
Ely Ronnen
Aaron Trudeau
Nick Maltbie
simensan
Michal Cichra
Paul Pacheco
Skjalg S. Mæhre
Robin
Paul Boocock
ivan-hernandez-scopely
mob-sakai
Webber Takken
David LeGare