[BUG] Markdown CI fails without retrievable job logs #6

New issue

Open

opened 2026-09-15 12:01:29 +02:00 by erikinkinen · 0 comments

erikinkinen commented

2026-09-15 12:01:29 +02:00

Owner

Summary

An older Posixd Markdown review CI run failed after roughly 61 minutes, and Forgejo cannot return its job log. The failed result has no complete diagnostic log, so a source, runner or infrastructure cause cannot yet be established. Current source validation and a later successful run do not replace that missing evidence.

Reproduction

Open Posixd run 10, API run 9287, job 19980, attempt 1. The selected signed commit is 16262cf5c339d46d447f20004a8bb6d6ff86adc6, triggered by review #5. The supported job-log request /repos/erix/posixd/actions/jobs/19980/logs?attempt=1 returns HTTP 500. Two terminal-state observations, separated by the enforced polling interval, retained the same failure and unavailable log.

Expected and actual behavior

Expected: a terminal job exposes a complete log that identifies its failure. Actual: run and job status are failure; the log endpoint returns HTTP 500 and no complete bytes can be classified. Started 2026-09-15 10:53:57 +02:00; stopped 11:54:52 +02:00; recorded duration 3655 seconds. The duration alone does not establish a timeout cause.

Environment and identity

Forgejo ci.yml, pull_request event, Markdown job on the docker runner class, task 18875. The same original source passed local Markdown/template validation. Later signed Posixd a174994ac6ebe5561f6bc1095cb3abcb3ec71d91 passes push/review runs 11/12 with complete warning-free logs. No Rust crate or guest runtime exists in this repository yet.

Evidence and prior checks

Both exact terminal snapshots and HTTP 500 dispositions are retained. No complete log was downloaded and no source regression, command failure or runner timeout is asserted without evidence. The prior component observation cohort is stopped with this failed/incomplete disposition; its other completed successful jobs retain their logs. No unchanged-input CI rerun was requested.

Tracking

Related to feature issue #1 and review #5. Restore or explain the missing job log, identify the failed step or infrastructure disposition, and record a corrective action before closing. Keep the current successful source checks and the older failed result distinct.

## Summary An older Posixd Markdown review CI run failed after roughly 61 minutes, and Forgejo cannot return its job log. The failed result has no complete diagnostic log, so a source, runner or infrastructure cause cannot yet be established. Current source validation and a later successful run do not replace that missing evidence. ## Reproduction Open [Posixd run 10](https://git.erikinkinen.fi/erix/posixd/actions/runs/10), API run 9287, job 19980, attempt 1. The selected signed commit is `16262cf5c339d46d447f20004a8bb6d6ff86adc6`, triggered by review #5. The supported job-log request `/repos/erix/posixd/actions/jobs/19980/logs?attempt=1` returns HTTP 500. Two terminal-state observations, separated by the enforced polling interval, retained the same failure and unavailable log. ## Expected and actual behavior Expected: a terminal job exposes a complete log that identifies its failure. Actual: run and job status are failure; the log endpoint returns HTTP 500 and no complete bytes can be classified. Started 2026-09-15 10:53:57 +02:00; stopped 11:54:52 +02:00; recorded duration 3655 seconds. The duration alone does not establish a timeout cause. ## Environment and identity Forgejo `ci.yml`, pull_request event, Markdown job on the docker runner class, task 18875. The same original source passed local Markdown/template validation. Later signed Posixd `a174994ac6ebe5561f6bc1095cb3abcb3ec71d91` passes push/review runs 11/12 with complete warning-free logs. No Rust crate or guest runtime exists in this repository yet. ## Evidence and prior checks Both exact terminal snapshots and HTTP 500 dispositions are retained. No complete log was downloaded and no source regression, command failure or runner timeout is asserted without evidence. The prior component observation cohort is stopped with this failed/incomplete disposition; its other completed successful jobs retain their logs. No unchanged-input CI rerun was requested. ## Tracking Related to feature issue #1 and [review #5](https://git.erikinkinen.fi/erix/posixd/pulls/5). Restore or explain the missing job log, identify the failed step or infrastructure disposition, and record a corrective action before closing. Keep the current successful source checks and the older failed result distinct.

erikinkinen added the

bug

phase-6

labels

2026-09-15 12:01:29 +02:00

erikinkinen referenced this issue

2026-09-15 12:02:38 +02:00

[FEATURE] Implement isolated compatibility realms and exact grant lifecycles #1

erikinkinen referenced this issue

2026-09-15 12:02:39 +02:00

WIP: Specify staged mediator custody and realm bootstrap ownership #5