- Fix issue with migrations not running automatically, due to pool size being too small
- Replace uglify-js with terser-js
- Use
nodev10.x or Docker images
- Update Elixir to 1.8, and update dependencies
- Upgrade to Phoenix 1.4
- Use
yarninstead ofnpmfor JS dependency management
- Manual build jobs can now be declared on the
.alloy-ci.ymlfile under thewhendirective. Manual jobs are those that need a user action before they can be picked up by a runner. Once previous stages of a pipeline have succeeded, a manual job becomes available for enqueuing.
- Handle Tentacat bug that causes repositories list to be returned in 2 different ways e144aeaa
- Allow users to re-authorize AlloyCI, so that they can see repositories from organizations with restricted permissions. See #52 for details on upcoming improvements.
- Use ETS to store intermediate build trace updates to avoid unnecessary round trips to the DB
- Update Elixir to 1.7, and update dependencies
- Replace Que with own implementation of background jobs using GenServer (since Que dependencies stopped working on Elixir 1.7)
- Added
typespecdefinitions to all public functions outside of thewebfolder - Updated
distilleryto 2.0.9 and removedmix_docker - Replaced
.buildand.releaseDockerfiles with a singleDockerfilethat builds in stages, and addedmaketasks to release and push the images - Replaced
elixir:latestwithelixir:slimfor the release stage in order to reduce final image size - Added CI task to automatically release Docker images when a tag is pushed
- MAJOR DEPRECATION JSON config files for build/pipeline configuration are no longer supported. Use YAML files instead. The main reason behind this decision is that YAML is much more flexible than JSON. It allows us to use aliases in oder to avoid duplication, to add comments to make the config file more understandable, and it is the de facto standard configuration file for almost all CI systems in the market.
- Build status is now automatically updated on the pipeline view. It uses websockets to redraw the HTML data, and update the view when the status changes.
- Pipeline status is also automatically updated in the same manner on the project view.
- Added a Makefile to facilitate the start up procedure in dev mode.
- Use YAML config files instead of JSON.
- Fixed redirect loop that would happen when an authentication token expires.
- Reverted back to Kerosene v0.7.0, as the updated v0.8.0 had pagination bugs.
- Added support for S3 compatible object storage to store build artifacts. See the documentation for instructions on how to set it up. Most S3 compatible services, like AWS S3, Digital Ocean Spaces, Minio, etc. should work.
- Improved runner ENV variables for better Coveralls support
- Fixed small bug on the
ArtifactSweeperGenServer. The server state was not being returned, which caused intermittent errors on subsequent runs (one would work, the next one would error out, the next one would work, etc.) - Properly populate the
$CI_COMMIT_REF_SLUGenvironment variable with the shorthand for the branch/tag name
- Updated dependencies
- Cleaned up data structures code
- Replaced use of
use AlloyCi.Web :schemawith properEctoimports
- Replaced use of
- Added support for GitHub Enterprise
- AlloyCI can now be used with a GitHub Enterprise installation. See the documentation for more info on how to set it up.
- Ditched core-ui in favor of a cleaner, more functional design that is more appealing to the eye
- Expired artifacts are now pruned once a day (interval can be configured)
- Artifacts can be kept forever, by clicking a button in the build page
- Build artifacts can now be downloaded from the build page
- For this, the build page was completely reworked
- It is now its own page with valuable information about the build, along the build trace and artifacts that may have been created
- Individual builds can now be restarted independently (build will be copied, and old data will be kept)
- When cancelling a pipeline, all builds were marked as cancelled, now only builds that are either
created,running, orpendingare updated. - After the upgrade to Guardian 1.0 it became impossible to add another authentication method to an existing user. This has been fixed.
- Show available global runners on the project's settings page
- Initial support for build artifacts:
- Jobs can create artifacts and upload them to the server (only local storage supported for now)
- Jobs can download artifacts from previous builds, if they depend on them
- Support for dependent builds:
- Builds with artifacts from previous stages will be automatically added as dependencies for jobs of the subsequent stage
- Builds can explicitly declare their own dependencies via the
.alloy-ci.ymlfile
- Added the option to use Sentry for error catching in production
- Updated Guardian to 1.0
- Redirect to main projects if user is logged in and visits the register page
- Skip updating the build trace, if the runner calls the endpoint with an empty string
- Updated JS dependencies
- Pipelines are now created when a pull request from a fork is submitted. This feature requires the newly released AlloyCI Runner (GitLab Runner fork), so it is incompatible with all previous and future GitLab Runner versions. (#29)
- Added support for
onlyandexceptkeywords for build jobs. This gives you better control over which jobs run when. (#14) - Regular users can now manage project specific runners for projects to which they have access. (#33)
- Projects now support the use of secret variables. They can be used to pass sensitive information to the runner without committing it to the repository. (#23)
- Redirect to the main projects' view when all auth options have already been created. (#30)
- Better processing of pipelines when the worker is called (5a9ca8f)
- Fix notification bug, that would send one on every failed build. (63dc787)
- All mentions of
integrationin the documentation, code, and configuration files have been replaced withapp. This is to conform with the changes imposed by GitHub and their API. (#28)
- First public release