fix(auto-apply): capture platformctl exit codes #381

Merged

pdurlej merged 1 commit from codex/fix/auto-apply-capture-exit-codes into main

2026-05-19 00:47:21 +02:00

codex commented

2026-05-19 00:45:57 +02:00

Collaborator

Canary status: missing — workflow guard fix; rely on required Forgejo checks before merge

Canary Context Pack

Product story

F3 one-shot smokes need the auto-apply workflow to reach the explicit guard when platformctl plan reports allowed drift with exit code 1. Today the shell exits before the guard runs, so minio-init cannot complete its designed backup-backed validation path.

What changed

Wrap platformctl plan, apply, and health invocations in if ...; then ec=0; else ec=$?; fi blocks.
Keep the existing policy exactly as-is: plan exit 1 is still only accepted when the stateful one-shot guard allows it.
Update the workflow contract test to require explicit exit-code capture.

Why it changed

Manual minio-init F3 dispatch was classified as eligible and produced the exact allowed drift (container.state.exited_success: exited 1 -> exited 0), but the workflow stopped immediately after platformctl plan and never ran allow_one_shot_stateful_apply.py.

Files touched

.forgejo/workflows/platformctl-auto-apply.yml
control-plane/platformctl/tests/test_forgejo_ci_scripts_contract.py

Runtime evidence

platformctl-auto-apply run #1446 / API id 1879: eligible minio-init, backup ref accepted, plan drift limited to container.state.exited_success, no apply executed.
home-platform-minio-1 remained healthy; home-platform-minio-init-1 remained the old exited(1) container.

Known constraints

This PR changes workflow control flow only. It does not broaden the stateful override, backup requirements, module allowlist, or host-agent command surface.

Explicit out-of-scope

No direct production mutation.
No MinIO bucket/config changes.
No change to backup policy or one-shot guard rules.

Requested decision

Merge, then rerun platformctl-auto-apply.yml for minio-init with allow_stateful=true, the existing MinIO backup ref, and BACKUP_DONE_F3.

Merge blockers

Any evidence that wrapping these commands weakens the guard. The code still checks and exits on the captured overall status.

Spec sources read

.forgejo/workflows/platformctl-auto-apply.yml — failing control-flow path.
control-plane/platformctl/ci/allow_one_shot_stateful_apply.py — guard semantics.
control-plane/platformctl/tests/test_forgejo_ci_scripts_contract.py — workflow contract coverage.
modules/minio-init/module.yaml and modules/minio-init/runbook.md — one-shot module contract.

Test plan

pytest -q control-plane/platformctl/tests/test_forgejo_ci_scripts_contract.py → 36 passed
git diff --check

Canary status: missing — workflow guard fix; rely on required Forgejo checks before merge ## Canary Context Pack ### Product story F3 one-shot smokes need the auto-apply workflow to reach the explicit guard when `platformctl plan` reports allowed drift with exit code 1. Today the shell exits before the guard runs, so `minio-init` cannot complete its designed backup-backed validation path. ### What changed - Wrap `platformctl plan`, `apply`, and `health` invocations in `if ...; then ec=0; else ec=$?; fi` blocks. - Keep the existing policy exactly as-is: plan exit `1` is still only accepted when the stateful one-shot guard allows it. - Update the workflow contract test to require explicit exit-code capture. ### Why it changed Manual `minio-init` F3 dispatch was classified as eligible and produced the exact allowed drift (`container.state.exited_success`: exited 1 -> exited 0), but the workflow stopped immediately after `platformctl plan` and never ran `allow_one_shot_stateful_apply.py`. ### Files touched - `.forgejo/workflows/platformctl-auto-apply.yml` - `control-plane/platformctl/tests/test_forgejo_ci_scripts_contract.py` ### Runtime evidence - `platformctl-auto-apply` run #1446 / API id 1879: eligible `minio-init`, backup ref accepted, plan drift limited to `container.state.exited_success`, no apply executed. - `home-platform-minio-1` remained healthy; `home-platform-minio-init-1` remained the old exited(1) container. ### Known constraints This PR changes workflow control flow only. It does not broaden the stateful override, backup requirements, module allowlist, or host-agent command surface. ### Explicit out-of-scope - No direct production mutation. - No MinIO bucket/config changes. - No change to backup policy or one-shot guard rules. ### Requested decision Merge, then rerun `platformctl-auto-apply.yml` for `minio-init` with `allow_stateful=true`, the existing MinIO backup ref, and `BACKUP_DONE_F3`. ### Merge blockers Any evidence that wrapping these commands weakens the guard. The code still checks and exits on the captured `overall` status. ## Spec sources read - `.forgejo/workflows/platformctl-auto-apply.yml` — failing control-flow path. - `control-plane/platformctl/ci/allow_one_shot_stateful_apply.py` — guard semantics. - `control-plane/platformctl/tests/test_forgejo_ci_scripts_contract.py` — workflow contract coverage. - `modules/minio-init/module.yaml` and `modules/minio-init/runbook.md` — one-shot module contract. ## Test plan - `pytest -q control-plane/platformctl/tests/test_forgejo_ci_scripts_contract.py` → 36 passed - `git diff --check`

codex added 1 commit

2026-05-19 00:45:57 +02:00

fix(auto-apply): capture platformctl exit codes

canary-required / collect-diff (pull_request) Successful in 3s

Details

platformctl plan / auto-apply scope (pull_request) Successful in 18s

Details

python-ci / Python 3.11 (pull_request) Successful in 37s

Details

python-ci / Python 3.12 (pull_request) Successful in 37s

Details

python-ci / Python 3.13 (pull_request) Successful in 37s