Skip to content

ci: allow the bun release-asset host in the fork review lanes - #11976

Merged
iamwhatever merged 1 commit into
mainfrom
fix/fork-lanes-release-assets-egress-11911
Sep 19, 2026
Merged

iamwhatever merged 1 commit into
mainfrom
fix/fork-lanes-release-assets-egress-11911

Conversation

@chenmingwei23

@chenmingwei23 chenmingwei23 commented Sep 19, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Every fork pull request has been unmergeable since 2026-09-18 21:25Z. All six fork review lanes finish review incomplete, and PR Readiness, the only required status on main, aggregates them. It is red on every fork PR, and re-running a lane does not clear it.

No lane reaches a verdict, and two different missing hosts are why. Five lanes die on a missing binary before any model call:

/home/runner/work/_temp/....sh: line 4: bun: command not found
##[error]Process completed with exit code 127

The GPT lane dies at the model call instead, with no bun in sight:

Sat, 19 Sep 2026 12:21:55 GMT:domain not allowed: bedrock-mantle.us-east-1.api.aws.
ERROR: Reconnecting... 1/5
ERROR: Connection failed: error sending request
##[warning]GPT 5.6 fork review pass 1 did not complete (exit 1); the verdict will fail closed.

Why it matters

No outside contributor can land anything, and the repository cannot merge its own repairs. #11888 is an already-approved fix for a red macOS shard, and it is stuck behind this.

A fork author has no way out from the branch either. Per #8445 the fork lanes consume no /ai-review override marker, so there is nothing to override.

Credit for the diagnosis is not ours. @buluoray found the bun release-asset mechanism in #11911. @rnoack1 found the second host on this PR: without it the fork GPT lane stays red after this merges.

What changed (motivation -> approach -> change)

claude-code-action installs bun before it runs anything, using oven-sh/setup-bun. That fetches bun as a GitHub release asset, and GitHub serves a release download by redirecting to release-assets.githubusercontent.com. That host is not objects.githubusercontent.com, which these allowlists already carried. All six lanes run step-security/harden-runner with egress-policy: block, so the download was refused, bun never installed, and the step exited 127 on the missing binary.

The GPT lane fails earlier and on a different host. Its two review passes run a CLI configured with model_provider = "amazon-bedrock", and that provider posts to https://bedrock-mantle.us-east-1.api.aws/openai/v1/responses. The allowlist carried the classic bedrock-runtime.* and bedrock.* hosts but not the mantle one, so both passes retried five times, exited 1, and the lane failed closed. Its claude-code-action use is only the Opus adjudication sub-step, which is why the bun signature never appears there.

Symptom six fork lanes review incomplete, PR Readiness red, no fork PR mergeable
Root cause A release-assets.githubusercontent.com blocked; bun's release asset is served there
Root cause B bedrock-mantle.us-east-1.api.aws blocked; the GPT passes call Bedrock there
Change allow each host, only in the jobs that make that call
Reach A: 6 jobs across 6 files. B: 1 job in 1 file. 3 other blocking-egress jobs left alone

Only the jobs that make each call get the endpoint. Five lanes have one bun job each. fork-security-scope-review.yml blocks egress in four jobs and runs the action in one, generate; its validate, adjudicate and publish jobs fetch nothing but Actions artifacts, so their allowlists are unchanged. The mantle host goes to the GPT lane alone, since the other five reach Bedrock through the runtime host.

Both hosts are named exactly, never as a wildcard. objects.githubusercontent.com is already allowed and carries the same class of GitHub-owned asset bytes, and the mantle host is the same Bedrock service the lane is already permitted to call, in the same region and under the same role.

The alternative for bun was installing it ahead of the action, which needs no allowlist change, since registry.npmjs.org is already allowed. It was rejected. The action pins its own bun-version internally, 1.3.14 today, and setup-bun skips its download only when the version already present matches that pin. A pre-installed bun works until the action bumps the pin, and then all six lanes break again the same way and just as silently. Allowing the host does not depend on a version.

Backwards compatibility

Compatible: this adds permitted destinations to fork egress allowlists and removes nothing. No input, field, type or contract is narrowed.

Tests

test/test_ai_review_workflows.py carries two classes, 19 tests, each pinning a coupling in both directions:

  • TestForkLaneBunEgress, parametrized over all six lanes: a job that runs claude-code-action behind egress-policy: block must allow the release-asset host, and a job that runs no model must not.
  • TestForkGptLaneMantleEgress: the job that configures the amazon-bedrock provider must allow the mantle host, and every lane job that configures no such provider must not.

All four assertions are mutation-checked. Dropping either endpoint fails the positive test for that host; adding the release-asset host to the security-scope validate job, or the mantle host to fork-opus-review.yml, fails the matching negative test; restoring returns 19 passed.

test/test_ai_review_workflows.py plus test/test_pr_quality_gates.py: 823 passed. flake8 and isort clean. mypy reports 8 errors in this file on this branch and the same 8 on origin/main, all at lines 15 to 9281, none inside the added code, which starts at line 11751.

Manual verification

Three facts were checked by hand, because these lanes cannot check themselves.

The bun download URL: requesting https://github.com/oven-sh/bun/releases/download/bun-v1.3.14/bun-linux-x64.zip answers 302 with location: https://release-assets.githubusercontent.com/.... That is the exact URL setup-bun logs.

The bun refusal: in Fork Opus 4.8 Review run 35422579467, harden-runner logs domain not allowed: release-assets.githubusercontent.com at 04:57:05, the same second the step logs its bun download. The same triple of download, bun: command not found and exit 127 appears in the design, UX and first-principles lanes.

The mantle refusal: in Fork GPT 5.6 Review run 35442578828, on fork PR #11027, harden-runner logs domain not allowed: bedrock-mantle.us-east-1.api.aws at 12:21:55. The discovery pass starts reconnecting at 12:21:59 and ends Connection failed: error sending request at 12:22:20; the falsification pass repeats it and the lane fails closed. That run logs no bun error at all.

One loose end, stated rather than hidden. That same run also blocked chatgpt.com in the same second, and this PR does not allow it. The CLI is pinned to model_provider = "amazon-bedrock", so its default ChatGPT backend should not carry a model call, and allowing a host on that suspicion would widen egress for nothing. If a fork GPT lane still fails after this merges with chatgpt.com as the only blocked host in its log, that host is the next line, and the log will say so plainly.

This PR's own board cannot prove any of it. workflow_run lanes always execute the default branch's copy of the workflow, so any lane that runs against this PR reads the old allowlist. A green board here is not evidence for the fix, and a red one is not evidence against it. Confirmation is the first fork PR to run a lane after this merges.

Related Issues

Fixes #11911

Pattern harvest

Rule candidate: review-prompt
Pattern: one symptom on a shared surface can have more than one cause. Five lanes and a sixth shared review incomplete, and the sixth needed a different host. Reading one failing lane's log and generalizing would have shipped a fix that left the outage half in place. Read one log per distinct lane shape.

Two suggestions beyond this fix.

First, a correction to the timing theory in #11911. The claude-code-action bump is probably not the trigger. fork-security-scope-review.yml is still on the older action pin, and that pin runs the same oven-sh/setup-bun at the same bun-version: 1.3.14 as the newer one the other five use. So the download predates the bump, and the change is either in GitHub's release-asset serving or in harden-runner enforcement. Which one is not isolated here.

Second, the evidence for this outage was sitting in every failed run. harden-runner's post-step prints domain not allowed: <host> and nothing reads it. A lane that fails while that line is present could name the host in its own summary, which would have turned a day of triage into one line, and would have caught the second host on day one.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable). N/A: docs/ci/ci-and-reviews.md describes these lanes as using "a narrow endpoint allowlist" and does not enumerate the hosts, so it does not go stale.
  • No secrets, credentials, or internal references in the diff

@chenmingwei23
chenmingwei23 requested review from a team and pepmach September 19, 2026 07:15
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Sep 19, 2026
@chenmingwei23

Copy link
Copy Markdown
Contributor Author

Intent: Let the six fork review lanes install bun again, by allowing the one GitHub host that serves bun's release asset, so PR Readiness stops failing closed on every fork PR.

Not a goal: changing any action pin, changing what the reviewers review, or answering the separate question of which upstream change made this start on 2026-09-18. This PR only widens each affected job's egress allowlist by that one host, and pins the coupling with a test.

@github-actions

github-actions Bot commented Sep 19, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Fable 5) — 🟡 CONCERNS

Design-level review of 87cf02d9b6c7ff069fd103e084c037b1b8b16dae — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: CONCERNS

Undocumented egress widening: the diff adds a second host, bedrock-mantle.us-east-1.api.aws, that the PR description never mentions, justifies, or verifies.

Watch

The description documents exactly one change ("Change | allow that one host", "Reach | 6 jobs across 6 files", tests "returns 12 passed"), but the diff also adds bedrock-mantle.us-east-1.api.aws:443 to fork-gpt-review.yml plus a second test class (TestForkGptLaneMantleEgress) — a separate fix for a separate failure, justified only in code comments. These allowlists are the egress boundary guarding review of untrusted fork content, so a new destination that reviewers reading the description will never examine is precisely the pattern this boundary exists to catch; the description's manual-verification section (302 check, harden-runner refusal log) covers the bun host only, leaving the mantle host's claim ("the URL a real job log shows") unevidenced. Either document and evidence it in the description or move it to its own PR.
Clears when: the description names the mantle endpoint, cites the run/log line showing the GPT lane calling it, and updates the test-count claim — or the mantle hunks move to a separate PR.

[DESIGN-REVIEWED] 87cf02d

@github-actions

github-actions Bot commented Sep 19, 2026 •

Copy link
Copy Markdown
Contributor

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of 87cf02d9b6c7ff069fd103e084c037b1b8b16dae and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 87cf02d

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 87cf02d9b6c7ff069fd103e084c037b1b8b16dae: <one-sentence reason>

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Sep 19, 2026
@github-actions

github-actions Bot commented Sep 19, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Fable 5) — 🟡 CONCERNS

Premise-level review of 87cf02d9b6c7ff069fd103e084c037b1b8b16dae — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

First-Principles-Verdict: CONCERNS

The description declares one host; the diff allows two — the bedrock-mantle egress entry on the credential-holding GPT job is undeclared and its defect unevidenced.

Not justified as shipped

  • Item 2 — undeclared, inherited: the description asserts "this adds one permitted destination to six egress allowlists" and "The host is named exactly", but the diff also adds bedrock-mantle.us-east-1.api.aws:443 to fork-gpt-review.yml — a second endpoint, on the job that holds Bedrock credentials — whose defect ("both passes retry five times… review incomplete") is asserted only by this PR's own comment and test docstring ("the URL a real job log shows"), with no run id or issue, unlike the bun fix's quoted run 35422579467. Grepped bedrock-mantle: 4 matches, all added by this diff; the base's codex-review.yml header ("via Amazon Bedrock (Mantle)") corroborates the provider name but not the host or the failure.
    Subtraction: defer — split the mantle host line, its comment block, and TestForkGptLaneMantleEgress into their own declared change.
    Clears when: a named run id or linked issue shows a GPT pass refused on bedrock-mantle.us-east-1.api.aws and the description declares the second host — or the mantle hunks are dropped from this PR.
  • Item 4 — rides along: TestForkGptLaneMantleEgress exists only to pin item 2's undeclared host.

What this change ships

Inventory (4 items) — 2 justified

Intent: unwedge fork PR merging by letting the six fork review lanes download bun through their blocking egress allowlists — a FIX. (Grepped egress-policy: block: 10 jobs; only the 6 edited ones run claude-code-action, so the fix has no unfixed sibling.)

  1. The six fork review lanes can install bun again — release-assets.githubusercontent.com allowed in each model job, non-model jobs left narrow — justified
  2. The GPT lane's allowlist additionally admits bedrock-mantle.us-east-1.api.aws — undeclared, inherited premise (defect asserted only inside the diff)
  3. New tests pin the bun-host ⇄ model-job coupling in both directions across all six lanes — justified
  4. New tests pin the mantle-host coupling on the GPT lane — rides along (with item 2)

[FIRST-PRINCIPLES-REVIEWED] 87cf02d

@github-actions

github-actions Bot commented Sep 19, 2026 •

Copy link
Copy Markdown
Contributor

Opus 4.8 Review — ✅ no blocking findings

Reviewed 87cf02d9b6c7ff069fd103e084c037b1b8b16dae — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 87cf02d

Verdict parsed from the review's SHA-scoped output markers for commit 87cf02d9b6c7ff069fd103e084c037b1b8b16dae.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 87cf02d9b6c7ff069fd103e084c037b1b8b16dae: <one-sentence reason>

@chenmingwei23
chenmingwei23 force-pushed the fix/fork-lanes-release-assets-egress-11911 branch from 231cfa7 to 4360de2 Compare September 19, 2026 07:23
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Sep 19, 2026
@rnoack1

rnoack1 commented Sep 19, 2026

Copy link
Copy Markdown
Contributor

One more host is needed alongside this, or the fork GPT 5.6 lane stays red after it merges: bedrock-mantle.us-east-1.api.aws:443 in .github/workflows/fork-gpt-review.yml.

To be clear about scope, since the four green lane comments above are the same-repo lanes on this in-repo branch: the breakage is only on the fork-*-review.yml path, where harden-runner runs egress-policy: block. On a fork PR, that lane logs

domain not allowed: bedrock-mantle.us-east-1.api.aws.

three seconds before the model call's first reconnect, then fails closed. Its allowlist carries the classic bedrock-runtime.*.amazonaws.com / bedrock.*.amazonaws.com hosts but not the api.aws mantle endpoint that #7706 documents the lane calling (https://bedrock-mantle.us-east-1.api.aws/openai/v1/responses).

It is GPT-specific — the fork Opus, Design and First-Principles lanes fail on release-assets.githubusercontent.com instead, which this PR does fix.

Since this PR already touches fork-gpt-review.yml and has no approvals yet, folding one line in here would avoid a second review cycle. The companion assertion in test_ai_review_workflows.py would want scoping to the GPT lane's model jobs, as the existing negative test asserts the host is absent from non-model jobs.

@pepmach — could you take the review? Every model-driven lane on every fork PR is currently reporting review incomplete, so no fork contribution is being reviewed at all until this lands.

Two hosts are missing from the fork lanes' blocking egress allowlists, and
each one takes a different set of lanes down.

claude-code-action runs oven-sh/setup-bun, which downloads bun from a
GitHub release asset. GitHub serves that from
release-assets.githubusercontent.com, a different host from the
objects.githubusercontent.com the six lanes already allow, so harden-runner
refused the download, bun never installed, and the reviewer step died
`bun: command not found` (exit 127). That is the Opus, Design, UX,
First-Principles and Security-Scope lanes.

The GPT lane breaks earlier and on a different host. Its two review passes
run a CLI whose amazon-bedrock provider posts to
bedrock-mantle.us-east-1.api.aws, which the allowlist's classic
bedrock-runtime and bedrock entries do not cover. harden-runner logs
`domain not allowed: bedrock-mantle.us-east-1.api.aws`, both passes then
retry five times and exit 1 with `Connection failed: error sending request`.

Either way the lane posts `review incomplete`. PR Readiness aggregates
those lanes, so no fork PR can reach a mergeable state.

Only the jobs that make each call get the endpoint.
fork-security-scope-review.yml blocks egress in four jobs and runs the
model in one, so its other three allowlists are unchanged, and the mantle
host goes to the GPT lane alone.

Tests pin both couplings in both directions, because workflow_run lanes
execute the default branch's yaml and cannot exercise this change.

Fixes #11911
@chenmingwei23
chenmingwei23 force-pushed the fix/fork-lanes-release-assets-egress-11911 branch from 4360de2 to 87cf02d Compare September 19, 2026 12:46
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: passed Eligible automated validation passed for the current revision labels Sep 19, 2026
@chenmingwei23

Copy link
Copy Markdown
Contributor Author

@rnoack1 Accepted and folded in. Head is now 87cf02d9b6.

I verified it independently before pushing rather than taking the host on trust:

So your scoping is right, and it also explains something I had left as an open discrepancy: the GPT lane was the one lane with no bun signature.

Two notes on how I implemented it.

The test is a second class, TestForkGptLaneMantleEgress, rather than an extra endpoint in the bun one. The two hosts have different owners: the bun host belongs to every job that runs claude-code-action, the mantle host to the job that configures the amazon-bedrock provider, and on this lane those are the same job for different reasons. Anchoring on the provider line keeps the negative direction meaningful, so the other five lanes must not carry the mantle host. Mutation-checked both ways: dropping it fails the positive test, adding it to fork-opus-review.yml fails the negative one.

That same run also blocked chatgpt.com in the same second, and I did not allow it. With the provider pinned to Bedrock its default ChatGPT backend should carry no model call, and I have no log line showing one, so allowing it would widen egress on a suspicion. If a fork GPT lane still fails after this merges with chatgpt.com as the only blocked host, that is the next line and its log will say so.

On folding it in here rather than a follow-up: this PR has no approvals to invalidate, it already touches that file, and shipping it alone would have left the outage half in place. The PR body now states both root causes, both failure signatures and the credit.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention labels Sep 19, 2026
@iamwhatever
iamwhatever merged commit 6406d42 into main Sep 19, 2026
130 of 133 checks passed
@iamwhatever
iamwhatever deleted the fix/fork-lanes-release-assets-egress-11911 branch September 19, 2026 17:35
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Sep 19, 2026
amadsalmon pushed a commit to amadsalmon/KiroCrew that referenced this pull request Sep 22, 2026
…tdev#11976)

Two hosts are missing from the fork lanes' blocking egress allowlists, and
each one takes a different set of lanes down.

claude-code-action runs oven-sh/setup-bun, which downloads bun from a
GitHub release asset. GitHub serves that from
release-assets.githubusercontent.com, a different host from the
objects.githubusercontent.com the six lanes already allow, so harden-runner
refused the download, bun never installed, and the reviewer step died
`bun: command not found` (exit 127). That is the Opus, Design, UX,
First-Principles and Security-Scope lanes.

The GPT lane breaks earlier and on a different host. Its two review passes
run a CLI whose amazon-bedrock provider posts to
bedrock-mantle.us-east-1.api.aws, which the allowlist's classic
bedrock-runtime and bedrock entries do not cover. harden-runner logs
`domain not allowed: bedrock-mantle.us-east-1.api.aws`, both passes then
retry five times and exit 1 with `Connection failed: error sending request`.

Either way the lane posts `review incomplete`. PR Readiness aggregates
those lanes, so no fork PR can reach a mergeable state.

Only the jobs that make each call get the endpoint.
fork-security-scope-review.yml blocks egress in four jobs and runs the
model in one, so its other three allowlists are unchanged, and the mantle
host goes to the GPT lane alone.

Tests pin both couplings in both directions, because workflow_run lanes
execute the default branch's yaml and cannot exercise this change.

Fixes kirodotdev#11911
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

ci: fork review lanes report "review incomplete" for every PR since ~22:42Z; model client cannot reach its backend

3 participants