Support having the Peril runtime be an AWS lamda job - orta
Move to a mono-repo - orta
You can opt out of checks support in your settings by adding "disable_github_check": true to your settings section. -
orta
BREAKING
This isn't breaking per-say, more that the default behavior for pull_request will now be more inline with how people
perceive it to act. Including myself.
Prior to this change; creating a new PR could trigger 1 to n webhooks depending on whether you've assigned reviewers,
assignees, added labels etc. Realistically, Peril shouldn't be running 5-6 times on the creation of a PR and that can
have simple side effects like duped comments, or do more work than you expect. So, this really simplifies the
responsibilities of pull_request and makes it just the key things that people are interested in.
If you need that functionality, use pull_request.* instead, and you'll get them all 👍
- orta (see #371)
Adds support for Sentry using if you set SENTRY_DSN in the env - orta
BREAKING
Replaces the scheduled task system with a fresh re-write and a new perspective now that I've had it running for a few months.
This issue covers what I want from the new scheduler (and why)
Basically if you're using the regular occurring scheduler I'd recommend reading the docs for the scheduler
but roughly, the keys in your "scheduler" now need to be one of the following: "hourly", "daily", "weekly",
"monday-morning-est", "tuesday-morning-est", "wednesday-morning-est", "thursday-morning-est",
"friday-morning-est".
There will be more keys in the future, but this covers all of my cases today.
- Fixes mis-attribution of the repo when an event's dangerfiles are between more than one repo - orta/barbosa
Migrated the repeated task scheduler to use agenda, this means it can take human readable strings, not just cron jobs - orta Decided to migrate schedulers to not allow free-form cron-like syntax - so expect breaking changes here in the future. - orta (see #361)
- All Peril run logs get sent to slack - orta
-
The edit an installation mutation allows setting a
installationSlackUpdateWebhookURLon your installation, which Peril will use to send messages to your installation - orta -
Generic rules for all the org now default to the settings JSON repo, reducing dupes in the settings json - orta
{
"$schema": "https://raw.githubusercontent.com/danger/peril/master/peril-settings-json.schema",
"rules": {
"create (ref_type == tag)": [
+ "org/new_tag.ts",
- "danger/peril-settings@org/new_tag.ts",
+ "org/updateDangerSystems.ts",
- "danger/peril-settings@org/updateDangerSystems.ts"
],
+ "pull_request.closed": "org/aeryn.ts",
- "pull_request.closed": "danger/peril-settings@org/aeryn.ts",
+ "pull_request": ["org/debug.ts", "org/changelog.ts"],
- "pull_request": ["danger/peril-settings@org/debug.ts", "danger/peril-settings@org/changelog.ts"],
+ "issue_comment": "org/markAsMergeOnGreen.ts",
- "issue_comment": "danger/peril-settings@org/markAsMergeOnGreen.ts",
+ "issues.opened": "org/checkTemplate.ts",
- "issues.opened": "danger/peril-settings@org/checkTemplate.ts",
+ "status.success": "org/mergeOnGreen.ts"
- "status.success": "danger/peril-settings@org/mergeOnGreen.ts"
},
"tasks": {
+ "trigger-scheduler": "testing/trigger-scheduler.ts",
- "trigger-scheduler": "danger/peril-settings@testing/trigger-scheduler.ts",
+ "logger": "testing/logger.ts"
- "logger": "danger/peril-settings@testing/logger.ts"
}
}I'll update thee docs later, once I'm certain all the edge cases are handled.
- Adds support for relative file imports inside a Dangerfile - orta
- Bug fix for recording webhooks, and updating the settings json from PR/pushes - orta
- Adds scheduler to staging - orta
- Fixes to the scheduled tasks feature - orta / MrCloud
- Improvements to the websockets information, all events now send names. - orta
- Running a task fixes - orta
- GitHubUtils now are now generated at runtime for event based PRs - orta
-
Adds a webhook system which allows admins to see:
- when a dangerfile is triggered
- the logs for a dangerfile
I'm still a little stuck on the client side (haven't figured how to run some code client-side with next) so it's still incubating, but the whole idea works. - orta
-
Adds apollo engine for debugging to the graphql api - orta
- Adds a new mutation for scheduling a task from a sandboxed run, where it doesn't have access to the agenda runtime - orta
- Adds a JSON schema for the settings json - orta
- Uses checks now - orta
[BREAKING CHANGES]
- Merged in Danger JS changes for multiple execution runs, and checks instead of comments you will need Checks API permissions, see the updated setup docs - orta
- Danger rules can use keypaths to determine if they should run - orta
- Added mutations for adding/editing/deleting env vars for hosted Peril - orta
- Hosted Peril can now have env vars in the DSL - orta
- Added mutations for re-sending a webhook through Peril - orta
- Fixed CORs support - orta
- Allows multiple inline events to trigger rules - SD10
- Adds initial support for recording all Webhooks going to an installation for 5 minutes - orta (idea from ashfurrow)
- Refactored a bunch of disparate event based routing into "plugins" still some better abstractions to go there but this is a good start - orta/pedrovereza
- The logs have been artistically hand crafted to feel good - orta
[BREAKING CHANGES]
-
Events, Tasks and Jobs all support Dangerfiles which have a default export function, in those cases the webhook/task data will be passed into the function as it's first argument. This is now the recommended way to get the event data.
Before:
// Stores the parameter in a closure that can be invoked in tests. const storeRFC = (reason: string, closure: () => void | Promise<any>) => // We return a closure here so that the (promise is resolved|closure is invoked) // during test time and not when we call rfc(). () => (closure instanceof Promise ? closure : Promise.resolve(closure())) // Either schedules the promise for execution via Danger, or invokes closure. const runRFC = (reason: string, closure: () => void | Promise<any>) => schedule(closure) const rfc: any = isJest ? storeRFC : runRFC // Note: Current WebHook for testing: 7ea07170-ee5e-11e7-827d-967e155710e3 // export const newTag = rfc("Send a comment to PRs on new tags that they have been released", async () => { const api = danger.github.api const gh = (danger.github as any) as Create
After:
export default async (create: Create) => {
This makes it:
- Trivial to Test, you can just import the file now
- Conceptually much simpler to think about. A PR Dangerfile still has the same DSL, but an event run is "just a function"
- Async work feels better now
- You can authenticate to Peril using 'Authenticate: Basic xxxyyzzz" with your JWT - orta
- Peril can redirect you to the GitHub app install page - orta
- Improvements to the GQL API - orta
- Created a real staging environment: https://staging-api.peril.systems/ - orta
- Peril is much more liberal about updating an installation's DB, PRs or pushes to the settings repo will trigger updates - orta
- Updated node to 9 - orta
- Does CHANGELOG checking on push - orta
- Adds a JWT auth system for API access to installations - orta
- Adds a GraphQL API for installation data - orta
- Consolidates the mongoDB representation, and the in-memory GitHubInstallation interfaces - orta
- Updated node to 8.4 - orta
- Updated Danger to 2.0a16 - orta
- Added the ability to skip specific repos - orta
Added the ability to expose an ENV var via the peril object in the runtime, I wanted to use the Slack API in one of
our Artsy Dangerfiles, but there wasn't a way to actually pass the secret to the Dangerfile. - orta