Repository navigation
Fix PHP downstream rebuild dispatch coverage and failures - #138
Conversation
|
Rollout prerequisite confirmed from the latest scheduled default-branch build: downstream job log shows an empty Before relying on this sender, configure/verify an appropriately scoped dispatch token for all five downstream repositories, and land their receivers first. This PR correctly makes missing-token failures visible, but does not create credentials or prove delivery/completion. Weekly receiver fallbacks remain important. No credentials were changed during this work. |
|
Correction to the earlier rollout note: a new DISPATCH_TOKEN is not normally required. This repository already inherits WORKFLOW_TOKEN; commit 5d40207 selects |
Make downstream dispatch failures visible and record outcomes without claiming downstream build completion.
PHP owns the weekly schedule and notifies WordPress, SQL Server, PHPQA and Swoole after publication. OCI8 is notified only by Swoole after all its parents are published; it is not a parallel direct PHP recipient. Downstream cron schedules are removed in their companion PRs.
Use an optional dedicated DISPATCH_TOKEN override, falling back to the existing inherited WORKFLOW_TOKEN. No new credential setup is normally needed; secret metadata cannot establish expiration or cross-repository permissions. If neither secret is available, fail visibly after image publication. No secret values or credentials were changed.
Also fix the template consistency check: the packaged fwd Blade cache cannot run as an arbitrary UID (reproduced locally); use the default container user. Regeneration leaves tracked image sources unchanged.
Related to #137. Merge downstream receivers first, then merge this sender and verify real dispatch/publication. PR validation does not establish that published derived images are patched.