Fix/runner self report crashes#50
Merged
Merged
Conversation
The nightly transformers triage was failing with "task runner exited without reporting (exit code 1)" — the runner crashed during startup (config/request build, checkout) before the LLM loop, and serge only saw the opaque exit code with no cause. - task_runner.run(): build config/request/clients *inside* the try so a startup exception self-reports a terminal error callback instead of exiting 1 silently; error now carries the exception + crashing frame. - task_runner.main(): backstop that POSTs a terminal error for anything that still escapes run()/run_review(). - webapp: fold collect_task_result's pod log tail into the reconcile error (was fetched then discarded) so the crash cause shows on the job. - webapp: TASK_DEBUG_MAX_TASKS caps /tasks per burst (debug knob) so a manual triage retrigger is a short crash-repro cycle, not the full batch. Set to 3 in prod.yaml temporarily; remove once diagnosed. - deploy: Infisical-backed secret sync (externalSecret) replacing local --secret-file, validated by this deploy. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The transformers-ci nightly triage opens a tracking issue then dispatches
one serge task per failure group, and refreshes the issue by polling open
PRs. Groups that end `no_fix` open no PR, so their row never updates — the
script has no way to learn a task's terminal status (it holds only a GitHub
Actions OIDC bearer, and /info requires a web session).
Add GET /tasks/{owner}/{repo}/{job_id}/status, authorized the same way as
POST /tasks (OIDC bearer, repo claim must match the path). Returns status /
result / error so the reconciler can mark no_fix (and error) groups in the
tracking issue instead of waiting on a PR that never comes.
Serge-side prerequisite only; the transformers-ci reconciler change to
consume it is a follow-up.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
tarekziade
force-pushed
the
fix/runner-self-report-crashes
branch
from
2f1850a to
c37bd8c
Compare
tarekziade
added a commit
that referenced
this pull request
Jul 6, 2026
* prod: drop TASK_DEBUG_MAX_TASKS cap so nightly triage runs the full batch The debug cap (TASK_DEBUG_MAX_TASKS=3) was added alongside the runner crash self-reporting (#50) to keep manual triage retriggers a short crash-repro cycle. Now that the runner-startup crash is diagnosed and self-reporting is in place, remove the temporary cap so serge executes every /tasks request in a burst again. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * ci: drop invalid trim() from ai-review workflow if-condition `trim()` is not a valid GitHub Actions expression function (the only string helpers are contains/startsWith/endsWith/format/join). Using it in the job `if:` made the workflow fail to compile — a startup failure with zero jobs run (e.g. run 28783571654). startsWith(comment.body, '@askserge ') already enforces "@askserge as the first word" since it matches from the literal start of the body, so dropping trim() keeps the #45 intent; we only lose tolerance for leading whitespace before the mention, which is an acceptable edge case. * ci: lint workflows with actionlint; fix trim() in docs example Add an `actionlint` job to tests.yml that the test/helm jobs depend on, so workflow files are statically checked on every PR. actionlint validates ${{ }} expressions (including that called functions exist), which catches errors like the invalid trim() that broke ai-review.yml — those otherwise surface only as a startup failure with zero jobs run. Also fix the same invalid trim() in the docs/github-action.md workflow example so users don't copy a workflow that fails to compile.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.