Conversation
Codex Review SummaryThis 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. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 797af48f0b
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".
| } | ||
| } | ||
| } | ||
| delay((result?.intervalSeconds ?: 30) * 1000) |
There was a problem hiding this comment.
When the first tracker successfully announces but returns no peers or only peers that fail metadata exchange, TrackerTiers records it as the current tracker and preferCurrentTracker() retries it first on every iteration. Since that tracker continues responding successfully, later supplied trackers are never queried; this delay therefore repeats until the overall metadata timeout, even if a fallback tracker has a usable peer. Rotate to the next tracker after exhausting the current response's peers.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Fixed in e2ca4ca. Tracker-only resolution now advances after exhausting each supplied tracker response, including empty responses and failed metadata peers. Metadata connections close and a bounded stopped announce finishes before the next tracker starts. The empty-response regression timed out before the fix; all tracker-only tests and the full JVM/iOS suites now pass. This commit also limits the inherited disk-failure test injection to one edit write so restarted checkpointing can succeed.
1 participant
The internal engine now accepts an explicit tracker-only privacy choice before magnet resolution. It requires validated supplied trackers, ignores explicit peers, avoids DHT, and tries tracker-returned metadata peers sequentially so old connections close before tracker failover. Public remains the existing default.
Hash-verified private metadata is accepted only on the tracker-only path. Public callers still reject it even on a cache hit. A task privacy field keeps public discovery disabled after public metadata is known: PEX is neither advertised nor accepted, and incoming peers require tracker-authorized hosts. Source/SDK selection and persisted privacy remain follow-up wiring; this does not expose a partially implemented public capability.
Stacked on #223; advances #162. Tests use real TCP metadata and peer-wire exchanges to cover authorized private resolution, unavailable tracker failover, ignored explicit peers, no DHT attempts, private-cache policy checks, missing/invalid trackers, cancellation without public fallback, PEX rejection, and incoming admission. Protocol references and remaining scope are documented in docs/design/torrent-v2-integrity.md.
Validation: 590 JVM and 569 executed iOS tests passed with zero failures/errors/skips, along with the native-runtime guard and two pinned v1 conformance scenarios. The stack includes #223's reviewed first-edit revision fix and CI test synchronization corrections.