Allow channel CI reruns to adopt promoted workflow fixes #112

Merged

gitchegumi merged 1 commit from kiyi/ci-reruns into main

2026-10-16 20:15:07 -05:00

kiyi commented

2026-10-16 19:21:18 -05:00

Member

Platform run 4998 attempt 2 was stopped by the generated stable-v1 first-attempt guard before source validation. Retrying after promoting a workflow fix is intended behavior.

Allow every positive Forgejo run attempt through the channel adapter while keeping its exact release checkout and workflow-channel.json release/run/attempt binding. Regression coverage retries the same owning run against a newer snapshot. Existing protected publication, Terraform and deployment actor guards remain unchanged; workflow names, job IDs, runner roles and required status names are preserved.

Validation: 17 channel tests, 6 candidate-publication tests and 8 native integration tests passed in a disposable Linux source checkout; Ruff F checks and git diff --check passed. Native Git Bash could not pass injected test environment variables, so the successful Linux results are the local evidence. Hosted rerun acceptance remains pending.

After merge, publish one workflow release and promote stable-v1, then retry Platform run 4998. Container startup retries are a separate runner-image change. Related: #93 and Homelab/platform#131.

Platform run 4998 attempt 2 was stopped by the generated stable-v1 first-attempt guard before source validation. Retrying after promoting a workflow fix is intended behavior. Allow every positive Forgejo run attempt through the channel adapter while keeping its exact release checkout and workflow-channel.json release/run/attempt binding. Regression coverage retries the same owning run against a newer snapshot. Existing protected publication, Terraform and deployment actor guards remain unchanged; workflow names, job IDs, runner roles and required status names are preserved. Validation: 17 channel tests, 6 candidate-publication tests and 8 native integration tests passed in a disposable Linux source checkout; Ruff F checks and git diff --check passed. Native Git Bash could not pass injected test environment variables, so the successful Linux results are the local evidence. Hosted rerun acceptance remains pending. After merge, publish one workflow release and promote stable-v1, then retry Platform run 4998. Container startup retries are a separate runner-image change. Related: #93 and Homelab/platform#131.

kiyi added 1 commit

2026-10-16 19:21:19 -05:00

Allow channel CI reruns to adopt promoted workflow fixes

Workflow acceptance / acceptance (pull_request) Successful in 1m43s

Details

Repository security / security-1 (pull_request) Successful in 4m5s

Details

Library checks / tests (pull_request) Successful in 6m44s

Details

b00f25a958

kiyi added this to the Homelab Roadmap project

2026-10-16 19:21:35 -05:00

portegumi approved these changes

2026-10-16 19:22:35 -05:00

portegumi left a comment

Reviewed the current head’s channel compiler, regression tests, and channel guidance. The adapter now accepts positive run attempts for stable-v1 or the snapshot’s exact version while retaining the immutable release-SHA library checkout and per-attempt workflow-channel.json binding. Tests cover valid reruns, invalid attempts/refs, and the same owning run with two releases; protected promotion, Terraform and deployment guards are not changed by this PR. No UDS route, admission or deployment-resource changes are introduced. Source approval only: hosted Workflow acceptance, Repository security, Library checks and an actual promoted-channel rerun remain unverified merge/operational gates. Nonblocking documentation follow-up: the Hosted acceptance paragraph still says to verify attempt-two rejection, and the batch-publishing paragraph still says an old failed attempt must not be rerun; align both with the new CI retry behavior, distinguishing protected operation reruns.

CI at submission: pending. This is a source-review approval, not a scan verdict; required checks still block merge.