Repository navigation
work: prepare current native fix packets through the existing App reader - #1338
oxa-agent[bot] wants to merge 18 commits into
Conversation
|
@codex review |
|
@codex security review |
Codex Review Summary: ❌ FailedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ce237a5dfe
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
🛡️ Codex Security Review
Here are some automated security review suggestions for this pull request.
Reviewed commit: ce237a5dfe
ℹ️ About Codex security reviews in GitHub
This is an experimental Codex feature. Security reviews are triggered when:
- You comment "@codex security review"
- A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review
Once complete, Codex will leave suggestions, or a comment if no findings are found.
|
@codex review |
|
@codex security review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7c394db943
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
🛡️ Codex Security Review
Here are some automated security review suggestions for this pull request.
Reviewed commit: 7c394db943
ℹ️ About Codex security reviews in GitHub
This is an experimental Codex feature. Security reviews are triggered when:
- You comment "@codex security review"
- A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review
Once complete, Codex will leave suggestions, or a comment if no findings are found.
|
@codex review |
|
@codex security review |
|
Codex Review: Didn't find any major issues. Delightful! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
There was a problem hiding this comment.
🛡️ Codex Security Review
Here are some automated security review suggestions for this pull request.
Reviewed commit: 3e0bd8c9d0
ℹ️ About Codex security reviews in GitHub
This is an experimental Codex feature. Security reviews are triggered when:
- You comment "@codex security review"
- A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review
Once complete, Codex will leave suggestions, or a comment if no findings are found.
| if let Some(model) = model { | ||
| packets = Input::new(&gh, &policy.host_path, repo, &token, &[], false) | ||
| .with_host(host) | ||
| .prepare_live(&claims, &board_policy, &worker, &partition, now, model)?; |
There was a problem hiding this comment.
🛡️ Codex Security Review
Security: Carry reviewed heads through native packet execution
With owner-enabled live mode and an eligible imported same-repository PR, this call publishes a fix packet whose retained binding has reviewed_head=Some(...). The native consumer reconstructs canonical bindings with NativeBinding::new, which always has reviewed_head=None, and rejects the mismatch before taking the packet. The reserved Claims lease then blocks that worker even after expiry, so a lower-privileged collaborator who can make such a PR eligible can stop later work until owner reconciliation. Preserve the reviewed head through packet parsing and add a refill-to-native-execution test.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Verified against the base native_task path: reconstructing a None-head canonical binding rejects the intended retained Some(reviewed_head). This runtime seam is owned by Lead in #1267; its actual published994b71a4 validates the full binding, derives Task.reviewed_pull_request from it, checks packet kind/head consistency, and retains the original admission through completion. The real owner fixture now uses new_reviewed with PR42/43 and retains the two-round/foreign-PR guards; Lead reports the composed hosted rounds pass. #1267 is currently still Draft/open, so I am retaining this dependency gate and keeping the protected rollout proposal disabled. No parallel execution implementation or current acceptance claim; refill-to-native composition must qualify against the actual landed/approved runtime source.
|
@codex review |
|
@codex security review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ad1037b604
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
🛡️ Codex Security Review
Here are some automated security review suggestions for this pull request.
Reviewed commit: ad1037b604
ℹ️ About Codex security reviews in GitHub
This is an experimental Codex feature. Security reviews are triggered when:
- You comment "@codex security review"
- A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review
Once complete, Codex will leave suggestions, or a comment if no findings are found.
|
@codex review |
|
@codex review |
|
@codex security review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b9a0ab1bf2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| if prepared.len() == 20 { | ||
| break; |
There was a problem hiding this comment.
Exclude already-published identities before applying the packet cap
If the first 20 ranked PRs have already-published packets but no remaining leases, and still have the same head/action, every live round prepares those same 20 packets and never reaches unpublished work ranked afterward. This can occur after an attempt finishes without changing the PR (for example, NothingToDo or Red) and its lease is released. Consumer::once uses a stable repository/number/head/action key, and helper_queue::publish_reserved rejects identities retained in pending or started storage, so all 20 packets are skipped indefinitely—even after the daily attempt counts reset. Exclude already-published identities before counting toward this cap, while retaining the publisher's locked deduplication check.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Verified on the actual source. Tests-first01bc9cfb518ec822ad5765ebaf76c52477c415cd, selected44299733978, reproduced this exact starvation at runtime on both architectures:1657 ran,1656 passed, only live_known_work_is_not_starved_by_twenty_published_identities failed, zero skips. Both failed tails show zero NEW packets versus one at line505. The fixture retains20 known lease-free publication identities in mixed plain/pinned ready, started and native-started layouts before eligible44, and preserves the earlier controls.
Lead owns only the existing helper_queue.rs dedupe extraction/read-only query on this same branch (handoff1308#6977874444). Work2 owns Input/Consumer/caller and the exact shared repo/number/head/action key. The prepared API takes the same canonical queue and exclusive spool lock, preserves pinned/plain/native checks, creates no started directory or reservation, and remains advisory; publisher repeats the checks while holding its lock through admission/publication. Actual published helper source and combined-head qualification are pending. No second store, fake publication probe, lease clearing, cap increase or new feature. This finding stays open until the exact fix qualifies; reviewed native execution remains independently open.
🛡️ Codex Security ReviewSecurity review completed. No security issues were found in this pull request. Reviewed commit: Only the user who started this review can view the report in Codex. ℹ️ About Codex security reviews in GitHubThis is an experimental Codex feature. Security reviews are triggered when:
Once complete, Codex will leave suggestions, or a comment if no findings are found. |
|
@codex review |
|
@codex security review |
|
Codex Review: Didn't find any major issues. 🎉 Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
There was a problem hiding this comment.
🛡️ Codex Security Review
Here are some automated security review suggestions for this pull request.
Reviewed commit: 01bc9cfb51
ℹ️ About Codex security reviews in GitHub
This is an experimental Codex feature. Security reviews are triggered when:
- You comment "@codex security review"
- A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review
Once complete, Codex will leave suggestions, or a comment if no findings are found.
| continue; | ||
| } | ||
| let fields = oxa_work::pull_request_fields(number)?; | ||
| let (snapshot, _) = self.reader.repository_read(&fields)?; |
There was a problem hiding this comment.
🛡️ Codex Security Review
Security: Check fork identity before fetching rich context
With owner-enabled live refill, an imported fix-eligible fork PR, and an outside contributor able to add enough comment/review text, this rich query can exceed the Reader's remaining 4 MiB budget before isCrossRepository is checked. repository_read then errors through ?, aborting the round and repeatedly starving later same-repository work. Fresh evidence beyond the earlier fork finding is that the new skip still occurs only after the entire rich response is captured. Fetch and validate minimal fork/head identity first, then request rich context only for same-repository PRs.
Useful? React with 👍 / 👎.
|
@codex review |
|
@codex security review |
|
Codex Review: Didn't find any major issues. 🎉 Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
🛡️ Codex Security ReviewSecurity review completed. No security issues were found in this pull request. Reviewed commit: Only the user who started this review can view the report in Codex. ℹ️ About Codex security reviews in GitHubThis is an experimental Codex feature. Security reviews are triggered when:
Once complete, Codex will leave suggestions, or a comment if no findings are found. |
|
@codex review |
|
@codex security review |
|
Codex Review: Didn't find any major issues. Delightful! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
🛡️ Codex Security ReviewSecurity review completed. No security issues were found in this pull request. Reviewed commit: Only the user who started this review can view the report in Codex. ℹ️ About Codex security reviews in GitHubThis is an experimental Codex feature. Security reviews are triggered when:
Once complete, Codex will leave suggestions, or a comment if no findings are found. |
|
@codex review |
|
@codex security review |
|
The account paying for this security review has reached its Codex usage limits. The payer can check the Codex usage dashboard. For personal accounts, using credits requires enabling “Use credits for security reviews” in Code review settings. If you do not manage the paying account, contact this repository's admins. |
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
T3: protected live queue publication for #1017 and the #307 assignment/lease and task-context requirements.
Owner-protected
live:trueconfiguration and an explicit model let the existing Input prepare at most 20 current, in-partition fix packets from the Claims-backed board. Consumer revalidates action/head and preserves hosted idle/bootstrap, existing-work, pause, original receipt, reviewed-head retention and locked hardlink publication fences. No new store, auth path or helper launcher.Preparation excludes unknown Claims rows, retained leases, unsupported native branches and already-published stable identities before counting toward the cap. Branch validation uses Work1's published canonical predicate (#1428, now merged). Standalone Markdown delimiter lines are quoted at the packet boundary; parser/publisher guards remain intact. Fork context currently skips after rich acquisition; the new budget finding below remains open.
Code4613181871 demonstrated starvation behind 20 already-published, lease-free identities. Tests-first01bc9cfb518ec822ad5765ebaf76c52477c415cd / lane44299733978 reproduced zero NEW packets instead of one on both architectures: 1657 ran, 1656 passed, only the new regression failed, zero skips. Prior mixed plain/pinned pending, legacy-started and native-started bytes and complete Claims rows must remain unchanged; those publication fixtures do not claim execution/completion.
Published repair951f7383dac82b2443ba6a7d7fcee897ab2c00eb incorporates Lead's exact helper-only f8db93e and uses its read-only locked
is_published(queue,key)before context/cap. Publisher repeats the same dedupe under its existing lock before reservation. The existing Consumer's identical key expression is shared; all other Consumer bytes, the native packet-take tail, retention/publication tail, Lead helper blob and before-regression test blob pass preservation checks. Exact-head selected 44303695498 passed 1,657 tests per architecture with zero skips. Full 44303704631 passed 5,059 tests once with zero runtime skips. Those runs tested 951f738. Earlier b9 lane44294159854 (1656 PASS per architecture, zero skips) and full44294169530 (5033 once, zero runtime skips) are prior evidence, not this head's qualification.New Security4613689313 is verified: rich fork context consumes the bounded Reader capture budget before fork identity is checked, allowing oversized context to abort the round before later same-repository work. Tests-first d083ea4 adds a real CLI case with fixed fork identity, a synthetic rich response exceeding 4 MiB, and eligible PR 44 afterward. Selected run 44305613811 reproduced the CLI capture refusal on x86_64 and ARM64. Each ran 1,658 tests, with 1,657 passing, only the new regression failing, and zero skips. This fixture exercises the transport limit with synthetic data. The prepared Input repair is preserved locally pending the coordinator’s disposition.
Current-head Code/Security, applicable read and normal merge queue remain gates. Reviewed native execution finding4610992922 remains open and depends on Lead #1267, now parked under Phase A; source availability and read-only composition are not execution acceptance. Deployment/protected config, authoritative imports/legacy exclusion, matching partition, full Goal/Session binding and #1264 publication hold remain independent. No live task, lease, model or helper was started. Phase A work is limited to verified findings on this existing branch.