This fixes cache propagation delay issues with Blacksmith's distributed cache.
The previous implementation used separate save/restore actions with
fail-on-cache-miss: true, which failed when the cache wasn't immediately
available after being saved by the Prepare job.
The new implementation uses actions/cache@v4 (combined save/restore)
with a fallback to do the checkout if cache miss, matching the pattern
used by cache-build and cache-db which don't have this issue.
Changes:
- Refactored cache-checkout action to use actions/cache@v4
- Added fallback checkout step when cache miss occurs
- Removed mode input (no longer needed)
- Updated all workflows to use the simplified cache-checkout action
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat: implement git checkout caching to speed up CI workflows
- Create cache-checkout action to save/restore git working directory
- Update prepare job in pr.yml to cache checkout after dangerous-git-checkout
- Update all downstream workflows to restore from cache instead of checkout
- Pass commit-sha to all workflow calls for cache key consistency
This reduces the number of full git checkouts from 20+ per workflow run to just 1,
significantly improving CI performance especially for large repos.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor: use actions/cache/restore directly instead of custom action
Remove sparse-checkout and use actions/cache/restore@v4 directly in all
downstream workflows. This eliminates the need for any git fetch operation
before restoring from cache, making the optimization more effective.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor: use github.event.pull_request.head.sha for cache key
- Remove commit-sha input from all downstream workflows
- Update cache-checkout action to use github.event.pull_request.head.sha directly
- Remove commit-sha output from prepare job
- Remove commit-sha from all workflow calls in pr.yml
- Keep sparse-checkout of .github to access the cache-checkout action
This eliminates the need to pass commit-sha around while still using
a reusable action for the cache restore logic.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix: update cache key to include branch name and add cache cleanup
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix: use github.head_ref and github.sha for cache key
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix: add prefix delete before saving cache to clear previous caches
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* perf: add sparse-checkout to exclude example-apps and mp4 files
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix: restore get_sha step and commit-sha output that was accidentally removed
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix: add trailing dash to cache key prefix to prevent accidental deletion of other branches' caches
The prefix-based cache deletion was using 'git-checkout-{branch}' which would
accidentally match and delete caches for branches with similar prefixes.
For example, branch 'feature' would delete caches for 'feature-2', 'feature-new', etc.
Adding a trailing '-' ensures exact branch matching:
- 'git-checkout-feature-' matches 'git-checkout-feature-abc123' (intended)
- 'git-checkout-feature-' does NOT match 'git-checkout-feature-2-def456' (correct)
Fixes both the cache-checkout action and the PR close cleanup workflow.
Co-Authored-By: unknown <>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* perf: optimize postgres health check in e2e workflow
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
* perf: optimize postgres health check in all e2e workflows
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore: migrate GitHub workflows from Buildjet to Blacksmith
- Replace buildjet-*vcpu-ubuntu-2204 runners with blacksmith-*vcpu-ubuntu-2204
- Replace buildjet/cache@v4 with actions/cache@v4
- Replace buildjet/setup-node@v4 with actions/setup-node@v4
- Replace buildjet/cache-delete@v1 with useblacksmith/cache-delete@v1
- Rename delete-buildjet-cache.yml to delete-blacksmith-cache.yml
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
* chore: bump Blacksmith runners to Ubuntu 24.04
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
* Reduce 16vcpu to 4vcpu for the API v2 E2E
* Remove 8vcpu usage
* chore: switch Blacksmith runners to ARM architecture
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
* chore: switch Blacksmith runners back to AMD (remove -arm suffix)
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
* fix: improve test cleanup to cover all bookings including reassignment-created ones
- Changed afterEach cleanup to find all bookings by eventTypeId instead of tracking bookingIds
- This ensures bookings created indirectly by managedEventManualReassignment are also cleaned up
- Removed problematic prefix-based deleteMany calls that could affect parallel tests
- Fixes idempotencyKey collision errors on high-parallelism CI runners
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
* fix: improve icons screenshot test stability with deviceScaleFactor and increased threshold
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
* fix: cap Playwright workers to 4 to match vCPU allocation on Blacksmith runners
Co-Authored-By: keith@cal.com <keithwillcode@gmail.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore: Remove unit tests from running after crowd in
* Update permissions on all workflows
* Updated pr permissions
* Added actions:write to top-level pr
* Update .github/workflows/unit-tests.yml
* Removed the linting annotation
* chore: Improve CI workflow
* Made a new pre-merge flow that runs the builds and e2e
* Refactored back into 1 file
* Adding another condition for checking the ready-for-e2e
* syntax
* More syntax
* Change e2e-required to only run when ready so it doesn't fail
* A different attempt at getting the ready-for-e2e label
* Supporting PRs and workflow_dispatch
* logging the branch
* hardcoding to test
* Trying the rest call
* Added logging
* Check for the label in the same step
* More logging
* Added job-level output
* Reverted back to 'true'
* Moved the e2e required check into the main required check
* More logging
* Peeling the branch off of the ref
* Fix the ref
* Trying a separate workflow for labeling
* Updated name for the labeled file
* Checking label events to avoid unnecessary runs
* Got rid of the extra check to see if required should run
* Upgraded old actions running Node 16
* Updated the pre-reqs for running jobs
* Removed the dependency of API builds before starting E2E suites
* Removed extra production build used by analyze job
* Fixed the run-e2e check
* Using the pull request object when available
* Checking for null PR
* Added check for no event
---------
Co-authored-by: Joe Au-Yeung <65426560+joeauyeung@users.noreply.github.com>
* Added cache retrieval steps for api v1, v2 in production builds workflow run
* Removed .turbo and .next specific arguments as path in v2 API build
---------
Co-authored-by: Keith Williams <keithwillcode@gmail.com>
* chore: Removed duplicate production build
* fix references
* Move env vars to top level of e2e
* job renaming
* Added env vars to top level of all e2e jobs
* Removed part of cache key that causes issues and is moot
* Using buildjet hardware so the caching works
* clean up
* chore: Add build of API v2 to PR checks
* Actually added the v2 build to the pipeline...
* Added back the original yarn install
* Updated the main PR flow to require API builds before tests