This PR makes the following changes:
- Adds an --id parameter to coder templates init so that you can non-interactively initialize a specific example template by ID (e.g. folder name)
- Updates develop.sh and lima/coder.yaml to use this parameter to select the docker example template.
- Deprecates the --experimental flag
- Adds a new flag --experiments which supports passing multiple comma-separated values or a wildcard value.
- Exposes a new endpoint /api/v2/experiments that returns the list of enabled experiments.
- Deprecates the field Features.Experimental in favour of this new API.
- Updates apidocgen to support type aliases (shoutout to @mtojek).
- Modifies apitypings to support generating slice types.
- Updates develop.sh to pass additional args after -- to $CODERD_SHIM.
The pid tracking refactor resulted in the pipe while echo to block the
script from continuing and showing the banner at the end. This change
redirects the stdout for the `start_cmd` to a new fd which `while` is
reading from.
* Use new `/healthz` endpoint for checking API liveness
* Improved credential handling/retrying in failure scenarios
* Separate site (`vite`) logs with prefix and date, additionally this
method also disables the `vite` clearing of the screen
* Show all interfaces coder API is listening on (due to `0.0.0.0`)
* Improved shutdown procedure / interrupt handling
* feat: Add tunnel by default
If an access URL is not specified, we will always tunnel.
This is from community-member feedback who exclaimed that
it's confusing having the default for `coder server` display
a warning message, and I agree.
There is very little (maybe none) in running `coder server`
without tunnel and without an access URL, so this seems like
overall a much better UX.
* Update install.sh
Co-authored-by: Ben Potter <ben@coder.com>
* Update docs/install/packages.md
Co-authored-by: Ben Potter <ben@coder.com>
* Fix reset pass test
* Fix e2e test
Co-authored-by: Ben Potter <ben@coder.com>
This was breaking the release process. Namely it was running
the `gen` targets due to the dependency tree, which was failing
on macOS and Linux runners. This revert can be reverted once
we fix that up.
This caused the following issues:
- Slim binaries weren't being updated.
- The coder.tar.ztd was misplaced.
- There is no coder.sha1 file with proper filenames.
This should be reintroduced in a future change with those fixes.
* fix: update agent ID in example templates
* fix: update agent ID in dogfood template
* chore: update default agent ID in documentation
* fix: develop.sh: start FE after template is created; leave template dir around if template creation fails
- Adds distinct exit statuses to the bootstrap scripts
- Makes the bootstrap scripts loop forever trying to download the coder agent
- Surfaces and logs the status codes returned by the download tool
This commit makes the following changes:
- Adds two variables docker_host and docker_arch to the example docker-code-server template
- Adds an example params.yaml to docker-code-server and updates the README.md to reference these parameters
- scripts/develop.sh will now attempt to create a template using docker-code-server with the appropriate parameters for the environment
- Updated Lima example to make use of the template parameters for docker-code-server
Additional drive-bys:
- webpack.dev.ts references CODER_HOST and not CODERV2_HOST; updated develop.sh accordingly
- develop.sh should now terminate child processes upon error.
- Running make dev now prompts you to run ./scripts/develop.sh manually, as GNU make does not appear to pass SIGINT to subprocesses.
- Added checks to develop.sh to ensure that coderd is listening before running our initial setup steps
- Add some more troubleshooting/debugging output to develop.sh
