feat(staged): centralize PR polling with tiered intervals and PR recovery#650
Open
feat(staged): centralize PR polling with tiered intervals and PR recovery#650
Conversation
Each BranchCardPrButton ran its own setInterval for PR status polling, leading to N concurrent gh CLI processes, stuck refreshing guards, and leaked event listeners. This replaces that with a single PrPollingService that coordinates all PR status refreshes via the existing backend refreshAllPrStatuses command. Key changes: - New PrPollingService with single timer, adaptive intervals, window focus awareness, concurrency protection, and stale-data tracking - Fix event listener race condition by awaiting listen() promises in cleanup instead of fire-and-forget .then() - Remove prStatusRefreshing guard, per-component setInterval, and window focus/blur handlers from BranchCardPrButton - Add stale-data indicator when polling fails repeatedly Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…component Add [PrPolling] prefixed console.info logs to help diagnose PR status update failures. Logs cover track/untrack, poll start/skip/end, scheduleNext intervals, refreshNow accept/drop, window focus/blur, pr-status-changed events, and $effect lifecycle in the component. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The create_pr and push_branch commands now emit a "running" session event (with branch_id, project_id, and session_type) before returning the session ID. This ensures the global sessionStatusListener registers the session before any completion event can arrive, fixing the race where fast-completing sessions hit "unknown session ID" warnings and PR numbers are never persisted. Remove the now-redundant sessionRegistry.register() and projectStateStore.addRunningSession() calls from the component's .then() callbacks. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When a branch has been pushed but has no prNumber (e.g. due to a race condition during PR creation), the BranchCardPrButton now checks GitHub for an existing open PR on mount via `gh pr view <branch>`. This runs async so the frontend isn't blocked. If a PR is found, the number is persisted and polling starts immediately. New backend command `recover_branch_pr` wraps the existing `get_pr_for_branch` GitHub CLI helper and updates the branch record. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
… component The info-level console.info logs were added to diagnose PR status update failures and are no longer needed. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The polling service now refreshes all projects, not just the one being viewed. The selected project polls at the existing 60s interval, while background projects poll every 5 minutes. Projects with pending CI checks still use the faster 15s interval regardless of selection. Key changes: - Replace branch-level track/untrack API with project-level setProjects and setSelectedProject, driven by reactive $effects in App.svelte - Per-project lastPolledAt tracking so each project polls independently at its own interval via a single adaptive timer - Remove component-driven track/untrack from BranchCardPrButton; stale notifications now keyed by projectId instead of branchId Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…anup Resolve code review feedback on the centralized PR polling service: - poll(): remove scheduleNext() from the early-return path when refreshInFlight is true — the in-flight operation's finally block already reschedules, avoiding a potential busy-wait cycle - setProjects(): short-circuit when the set of project IDs hasn't changed, preventing unnecessary poll triggers from Svelte $effect re-fires - refreshNow(): queue the projectId when a refresh is already in-flight instead of silently dropping it, so post-PR-creation and post-push refreshes are never lost - Stale notification: use exact equality (=== MAX_CONSECUTIVE_FAILURES) so the callback fires only on the transition to stale, not every subsequent failure cycle - Move unsubStale from onMount/onDestroy to $effect with cleanup, making it immune to double-mount (HMR, keyed re-render) and automatically reactive to branch.projectId changes - Add shouldAttemptRecovery() guard in the polling service to prevent N concurrent `gh pr view` CLI calls when many BranchCardPrButton components mount simultaneously for branches without PR numbers
![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: c7abde0b8d
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Comment on lines
+159
to
+165
| session_runner::emit_session_running( | ||
| &app_handle, | ||
| &session.id, | ||
| &branch_id, | ||
| &branch.project_id, | ||
| "pr", | ||
| ); |
There was a problem hiding this comment.
This emits a running session event before start_session(...) can fail, so if startup fails (for example provider/driver initialization errors), the command returns an error but the frontend has already registered the session as running via the global listener and may never receive a terminal event to clean it up. That leaves project/session UI stuck in a running state for PR creation (and the same ordering is repeated in the push path below).
Useful? React with 👍 / 👎.
Comment on lines
+123
to
+124
| refreshInFlight = false; | ||
| scheduleNext(); |
There was a problem hiding this comment.
refreshNow() queues pendingRefreshProjectId whenever refreshInFlight is true, but this path only gets drained in refreshNow(...).finally. When the in-flight operation is poll(), execution reaches this block and schedules the next timer without consuming the queued project, so immediate refresh requests that happen during a scheduled poll are silently deferred until the normal interval.
Useful? React with 👍 / 👎.
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.
Summary
BranchCardPrButtonwith a centralizedprPollingServicethat polls all projects app-wide with tiered intervals (15s for pending CI, 60s for selected project, 5min for background projects)recover_branch_prbackend command to look up existing open PRs on GitHub for branches missing PR numbers, with frontend coordination to prevent concurrentgh pr viewcallscreate_prandpush_branchcommandsTest plan
gh pr viewcalls when multiple branch cards mount simultaneously🤖 Generated with Claude Code