Playback performance and sync by richiemcilroy · Pull Request #1595 · CapSoftware/Cap

richiemcilroy · 2026-02-13T20:23:27Z

Refactor audio playback to a low-latency streaming path, optimize seeking logic, and reduce decoder initialization overhead.

This addresses user reports of audio lag and delayed startup, improves scrubbing responsiveness, and aims for smoother 60fps playback with better A/V synchronization across all supported platforms.

Note

High Risk
Touches core playback/audio timing, seek handling, and prefetch concurrency; regressions could manifest as A/V desync, stalls, or incorrect frames under rapid seeking across platforms.

Overview
Editor playback now supports live seeking while playing: Tauri seek_to/set_playhead_position only emit state on real changes and forward seeks to the active PlaybackHandle, while the timeline UI batches/coaleces drag seeks via requestAnimationFrame to avoid command storms.

Playback runtime is refactored to add a seek channel with generation-based invalidation, keyed BTreeMap prefetch buffering, dynamic FPS-scaled prefetch/timeout/skip tuning, and added startup/seek telemetry; audio playback defaults to the streaming buffer path with prerendered mode gated by CAP_AUDIO_PRERENDER_PLAYBACK, and AudioPlaybackBuffer is enabled cross-platform (incl. Windows).

Benchmarking is expanded with new scrub + startup metrics, optional JSON output for playback-test-runner and decode-benchmark (incl. fragmented input support), plus new docs/runbook and scripts to aggregate, validate, finalize, publish, analyze, and baseline-compare playback benchmark matrix results.

^{Written by Cursor Bugbot for commit 0cbac24. This will update automatically on new commits. Configure here.}

Greptile Overview

Greptile Summary

Refactors playback to use live seeking without stop/restart, switching from VecDeque to BTreeMap prefetch buffer with generation-based concurrency control to discard stale frames after seeks. Audio switches to low-latency streaming by default (prerender mode behind CAP_AUDIO_PRERENDER_PLAYBACK env var). Adds dynamic prefetch window tuning based on FPS, seek coalescing in Timeline UI via RAF batching, and telemetry for first-render and seek-settle latency.

Key improvements:

Eliminated playback stop/restart on seek - PlaybackHandle.seek() updates running loop via watch channel
Generation tagging prevents race where stale prefetched frames from old seek positions corrupt playback
Frontend Timeline coalesces rapid seeks with scheduleSeek + RAF to reduce IPC churn
Tauri commands avoid no-op state emissions and forward seeks directly to active handle

Issues found:

Lock acquisition failures in is_in_flight check (playback.rs:795) silently return false, triggering redundant decode
Timeline seek deduplication (index.tsx:284) blocks retry after failed seek because lastCompletedSeekFrame only updates on success
Audio resampler reset() (audio.rs:441) swallows errors, leaving stale state that causes sync drift
Dynamic skip threshold reduction during sustained lag may over-skip keyframes, worsening recovery

Confidence Score: 2/5

High risk - core playback timing and concurrency changes with multiple race conditions and error-swallowing paths that can cause A/V desync or frame corruption under load
Score reflects three critical logic bugs (redundant decode race, seek retry blocking, audio reset failure) plus complex generation-based concurrency that hasn't been battle-tested across platforms. The refactor touches ~800 lines in playback loop with new BTreeMap buffer, dual in-flight tracking, and dynamic threshold tuning - all high-risk areas for regressions in frame correctness and A/V sync
Pay close attention to crates/editor/src/playback.rs (concurrency/race conditions), apps/desktop/src/routes/editor/Timeline/index.tsx (seek retry logic), and crates/editor/src/audio.rs (error handling)

Important Files Changed

Filename	Overview
crates/editor/src/playback.rs	Major refactor adds live seeking with generation-based concurrency, dynamic prefetch tuning, BTreeMap buffer, and telemetry; high complexity with race condition risks
crates/editor/src/audio.rs	Adds Windows-specific `set_playhead_smooth` drift tolerance and safer `reset()` error handling; low-risk audio changes
apps/desktop/src-tauri/src/lib.rs	Optimizes `seek_to` and `set_playhead_position` to avoid no-op state emissions and forward seeks to active PlaybackHandle
apps/desktop/src/routes/editor/Timeline/index.tsx	Adds RAF-based seek coalescing and removes stop/restart on seek during playback; simpler scrubbing path with de-duplication

Sequence Diagram

sequenceDiagram
    participant UI as Timeline UI
    participant Tauri as Tauri Commands
    participant PH as PlaybackHandle
    participant PL as Playback Loop
    participant PF as Prefetch Task
    participant AU as Audio Thread
    
    UI->>Tauri: seekTo(frame)
    Tauri->>Tauri: Check state changed
    alt State Changed
        Tauri->>Tauri: Emit state update
        Tauri->>PH: seek(frame)
        PH->>PL: seek_rx.send(frame)
        PL->>PL: Increment seek_generation
        PL->>PL: Clear prefetch_buffer & frame_cache
        PL->>PF: Send new generation via seek_generation_tx
        PL->>AU: Update audio_playhead_tx
        PF->>PF: Drop stale frames (old generation)
        PF->>PF: Reset prefetch from new frame
        PF->>PL: Send new frames with generation tag
        PL->>PL: Filter by generation, render frames
    end
    
    Note over UI,AU: Live seeking without stop/restart
    
    UI->>UI: scheduleSeek coalescing
    UI->>Tauri: Batched IPC call
    Note over PL,PF: Generation-aware concurrency prevents stale frames

_{Last reviewed commit: 0cbac24}

cursor · 2026-02-13T20:23:27Z

Cursor Agent can help with this pull request. Just @cursor in comments and I'll start working on changes in this branch.
_{Learn more about Cursor Agents}

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

…ithub.com/CapSoftware/Cap into cursor/playback-performance-and-sync-dec3

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

…ithub.com/CapSoftware/Cap into cursor/playback-performance-and-sync-dec3

tembo · 2026-02-14T00:19:08Z

crates/editor/src/playback.rs

                                }
                            }

+                            if buffered_wait_prefetch_changed {


Small perf nit: if a seek came in during the wait loop, you continue right after this and the seek handler clears prefetch_buffer anyway, so you can bail before trimming.

Suggested change

if buffered_wait_prefetch_changed {

if seek_rx.has_changed().unwrap_or(false) {

continue;

}

if buffered_wait_prefetch_changed {

trim_prefetch_buffer(&mut prefetch_buffer, frame_number);

}

cursor

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

This PR is being reviewed by Cursor Bugbot

Details

You are on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.

cursor · 2026-02-14T00:19:43Z

apps/desktop/src/routes/editor/Timeline/index.tsx

+			frameNumber === pendingSeekFrame ||
+			frameNumber === inFlightSeekFrame ||
+			frameNumber === lastCompletedSeekFrame
+		) {


Seek dedupe blocks valid repeat seeks

Medium Severity

scheduleSeek drops requests when frameNumber === lastCompletedSeekFrame. Because lastCompletedSeekFrame is never cleared on normal playback progress, a later seek back to a previously completed frame can be ignored even after the playhead moved away. This makes timeline scrubbing intermittently no-op in Timeline/index.tsx.

Additional Locations (1)

apps/desktop/src/routes/editor/Timeline/index.tsx#L310-L311

cursor · 2026-02-14T00:19:43Z

crates/editor/src/playback.rs

+                *current_frame = frame_number;
+                true
+            }
+        });


Repeated same-frame seeks are ignored

Medium Severity

PlaybackHandle::seek uses seek_tx.send_if_modified, so a seek to the same frame as the last requested seek is dropped. Because seek_tx is never updated as playback advances, later valid seeks back to that frame can be ignored even after the playhead moved away.

cursor · 2026-02-14T00:19:43Z

apps/desktop/src/routes/editor/Timeline/index.tsx

+		if (zoomRafId !== null) cancelAnimationFrame(zoomRafId);
+		if (scrollRafId !== null) cancelAnimationFrame(scrollRafId);
+		if (seekRafId !== null) cancelAnimationFrame(seekRafId);
+	});


Unmount doesn’t cancel pending seek pipeline

Low Severity

onCleanup only cancels current RAF IDs, but it does not invalidate flushPendingSeek while an async commands.seekTo is in flight. After unmount, the finally block can schedule another RAF and continue issuing seeks from a disposed Timeline instance.

Additional Locations (1)

apps/desktop/src/routes/editor/Timeline/index.tsx#L314-L319

cursor

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

This PR is being reviewed by Cursor Bugbot

Details

You are on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.

cursor · 2026-02-14T00:20:31Z

apps/desktop/src-tauri/src/lib.rs

+
+    if let Some(handle) = playback_handle {
+        handle.seek(frame_number);
+    }


Seek dedupe uses stale state

Medium Severity

seek_to and set_playhead_position now skip handle.seek(...) when state.playhead_position already equals frame_number. But playhead_position is not advanced by playback frames, so it can be stale. This drops valid seeks to a previously requested frame while playback has moved, making scrubbing/rewind requests intermittently ignored.

Additional Locations (1)

apps/desktop/src-tauri/src/lib.rs#L1917-L1940

cursor · 2026-02-14T00:20:31Z

crates/editor/examples/decode-benchmark.rs

+        }
+    }
+
+    if let Err(error) = fs::write(&combined_path, &combined_data) {


Fragment merge can exhaust memory

Medium Severity

get_fragmented_video_duration reads init.mp4 and every .m4s segment into one in-memory Vec before writing a temp file. Large fragmented recordings can allocate very large buffers and terminate the benchmark process with OOM, since there is no size guard or streaming append path.

greptile-apps

_{21 files reviewed, 6 comments}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-02-14T00:21:15Z

crates/editor/src/playback.rs

                            prefetched.segment_index,
                        ))
                    } else {
-                        let is_in_flight = main_in_flight
+                        let in_flight_key = (seek_generation, frame_number);
+                        let is_in_flight = playback_prefetch_in_flight
                            .read()
-                            .map(|guard| guard.contains(&frame_number))
-                            .unwrap_or(false);
+                            .map(|guard| guard.contains(&in_flight_key))
+                            .unwrap_or(false)
+                            || playback_decode_in_flight
+                                .read()
+                                .map(|guard| guard.contains(&in_flight_key))


both playback_prefetch_in_flight and playback_decode_in_flight locks checked separately with .unwrap_or(false) fallback - if lock acquisition fails, returns false and playback starts redundant decode even though frame may already be in-flight in prefetch task

Prompt To Fix With AI

This is a comment left during a code review. Path: crates/editor/src/playback.rs Line: 795:805 Comment: both `playback_prefetch_in_flight` and `playback_decode_in_flight` locks checked separately with `.unwrap_or(false)` fallback - if lock acquisition fails, returns `false` and playback starts redundant decode even though frame may already be in-flight in prefetch task How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2026-02-14T00:21:16Z

apps/desktop/src/routes/editor/Timeline/index.tsx

+	function scheduleSeek(frameNumber: number) {
+		if (
+			frameNumber === pendingSeekFrame ||
+			frameNumber === inFlightSeekFrame ||
+			frameNumber === lastCompletedSeekFrame
+		) {
+			return;
+		}
+		pendingSeekFrame = frameNumber;
+		if (seekRafId === null) {
+			seekRafId = requestAnimationFrame(flushPendingSeek);
+		}
+	}


deduplicates seeks against lastCompletedSeekFrame which is only set on success (line 311) - if seekTo throws (line 313), lastCompletedSeekFrame stays stale, and retrying the same frame later gets incorrectly rejected as duplicate on line 284

Prompt To Fix With AI

This is a comment left during a code review. Path: apps/desktop/src/routes/editor/Timeline/index.tsx Line: 280:292 Comment: deduplicates seeks against `lastCompletedSeekFrame` which is only set on success (line 311) - if `seekTo` throws (line 313), `lastCompletedSeekFrame` stays stale, and retrying the same frame later gets incorrectly rejected as duplicate on line 284 How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2026-02-14T00:21:17Z

crates/editor/src/audio.rs

    pub fn reset(&mut self) {
-        *self = Self::new(self.output).unwrap();
+        if let Ok(resampler) = Self::new(self.output) {
+            *self = resampler;
+        }
    }


Self::new() can fail and return Err, but error is silently ignored - resampler stays in old state, causing audio sync drift or corruption after seek when sample rate/channel config changes

Prompt To Fix With AI

This is a comment left during a code review. Path: crates/editor/src/audio.rs Line: 439:443 Comment: `Self::new()` can fail and return `Err`, but error is silently ignored - resampler stays in old state, causing audio sync drift or corruption after seek when sample rate/channel config changes How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2026-02-14T00:21:18Z

crates/editor/src/playback.rs

+                        scheduled_in_flight_frames.insert(frame_num);
                        if let Ok(mut in_flight_guard) = prefetch_in_flight.write() {
-                            in_flight_guard.insert(frame_num);
+                            in_flight_guard.insert((generation, frame_num));
                        }


tracked in both local scheduled_in_flight_frames HashSet (line 413) and shared prefetch_in_flight RwLock (line 415) - consider using only the local set to reduce contention, since prefetch task already removes from shared set on completion

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

Prompt To Fix With AI

This is a comment left during a code review. Path: crates/editor/src/playback.rs Line: 413:416 Comment: tracked in both local `scheduled_in_flight_frames` HashSet (line 413) and shared `prefetch_in_flight` RwLock (line 415) - consider using only the local set to reduce contention, since prefetch task already removes from shared set on completion <sub>Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!</sub> How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2026-02-14T00:21:19Z

apps/desktop/src-tauri/src/lib.rs

+    let state_changed = {
+        let state = editor_instance.state.lock().await;
+        state.playhead_position != frame_number
+    };
+
+    if state_changed {
+        editor_instance
+            .modify_and_emit_state(|state| {
+                state.playhead_position = frame_number;
+            })
+            .await;
+    }


locks state twice (lines 1919, 1932) - second lock held during seek() call which may briefly block - consider reading both playhead_position and playback_task in single lock scope to reduce hold time

Prompt To Fix With AI

This is a comment left during a code review. Path: apps/desktop/src-tauri/src/lib.rs Line: 1918:1929 Comment: locks `state` twice (lines 1919, 1932) - second lock held during `seek()` call which may briefly block - consider reading both `playhead_position` and `playback_task` in single lock scope to reduce hold time How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2026-02-14T00:21:20Z

crates/editor/src/playback.rs

@@ -724,10 +1093,29 @@ impl Playback {
                        {


dynamic_skip_threshold reduced progressively during sustained lag (late_streak) - this aggressively drops frames when decoder is slow, but may skip over keyframes needed for next GOP, potentially causing further decode delays instead of recovery

Prompt To Fix With AI

This is a comment left during a code review. Path: crates/editor/src/playback.rs Line: 1073:1093 Comment: `dynamic_skip_threshold` reduced progressively during sustained lag (`late_streak`) - this aggressively drops frames when decoder is slow, but may skip over keyframes needed for next GOP, potentially causing further decode delays instead of recovery How can I resolve this? If you propose a fix, please make it concise.

richiemcilroy and others added 29 commits February 13, 2026 22:26

improve: use keyed prefetch buffer for faster frame lookup

e1fb5e4

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: prune stale prefetched frames using ordered map

3eacadb

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: prune stale prefetched frames using ordered map

9980169

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: include comparison gate status in published matrix summaries

bde144b

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: include comparison gate status in published matrix summaries

f2efab8

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: tighten warmup timing and skip pruning in keyed prefetch path

576f1fd

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: tighten warmup timing and skip pruning in keyed prefetch path

634e18f

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: report candidate-only coverage in benchmark comparisons

ab756a9

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: report candidate-only coverage in benchmark comparisons

a0e2a17

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: include candidate-only count in published comparison summary

0a19cdb

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: include candidate-only count in published comparison summary

16f7170

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: add strict candidate-only gating for matrix comparisons

53f21e1

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: add strict candidate-only gating for matrix comparisons

7b8a79d

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: require contiguous prefetch coverage for warmup readiness

0202484

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: require contiguous prefetch coverage for warmup readiness

9f94a26

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: surface comparison coverage policies in published summaries

b7652b7

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: surface comparison coverage policies in published summaries

b1b534e

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: emit finalize summary json artifacts for automation

af068d9

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: emit finalize summary json artifacts for automation

a7ad109

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: optimize contiguous prefetch warmup scan

d6e0f6d

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: optimize contiguous prefetch warmup scan

e0dc3ef

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: include git metadata in finalize summary artifacts

5fd803b

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: include git metadata in finalize summary artifacts

91055a8

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: publish finalize summary metadata in matrix reports

ba37061

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: publish finalize summary metadata in matrix reports

c1e1f85

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: aggregate multi-run comparison metrics by key

1434554

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: aggregate multi-run comparison metrics by key

c332ada

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: defer contiguous warmup scans until first frame arrival

19b0083

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: defer contiguous warmup scans until first frame arrival

efc29ab

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

richiemcilroy and others added 23 commits February 14, 2026 00:14

improve: short-circuit frame waits when seeks are pending

d53082b

improve: short-circuit frame waits when seeks are pending

ee3ee2d

improve: short-circuit frame waits when seeks are pending

5e8ffd6

improve: skip stale waits and decode fallback on pending seeks

51044cf

improve: skip stale waits and decode fallback on pending seeks

7f796ff

improve: skip stale waits and decode fallback on pending seeks

3c81ece

improve: batch prefetch buffer trims during queue drains

75892b7

improve: batch prefetch buffer trims during queue drains

09e8307

improve: batch prefetch buffer trims during queue drains

dad0f79

improve: preserve prefetch ramp on small forward rebases

94fdebe

improve: preserve prefetch ramp on small forward rebases

09bb2f0

improve: preserve prefetch ramp on small forward rebases

6d0b272

improve: gate behind-prefetch scans per playback frame

6b4f0c9

improve: gate behind-prefetch scans per playback frame

4beb9af

improve: gate behind-prefetch scans per playback frame

c6e1116

Merge branch 'cursor/playback-performance-and-sync-dec3' of https://g…

d632a0f

…ithub.com/CapSoftware/Cap into cursor/playback-performance-and-sync-dec3

Merge branch 'cursor/playback-performance-and-sync-dec3' of https://g…

83cfbad

…ithub.com/CapSoftware/Cap into cursor/playback-performance-and-sync-dec3

Merge branch 'cursor/playback-performance-and-sync-dec3' of https://g…

2788856

…ithub.com/CapSoftware/Cap into cursor/playback-performance-and-sync-dec3

Merge branch 'cursor/playback-performance-and-sync-dec3' of https://g…

fbcd20e

…ithub.com/CapSoftware/Cap into cursor/playback-performance-and-sync-dec3

Merge branch 'cursor/playback-performance-and-sync-dec3' of https://g…

6e05540

…ithub.com/CapSoftware/Cap into cursor/playback-performance-and-sync-dec3

improve: batch prefetch trims in in-flight wait buffering

977414e

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

improve: batch prefetch trims in in-flight wait buffering

1fea346

Co-authored-by: Richie McIlroy <richiemcilroy@users.noreply.github.com>

Merge branch 'cursor/playback-performance-and-sync-dec3' of https://g…

0cbac24

…ithub.com/CapSoftware/Cap into cursor/playback-performance-and-sync-dec3

richiemcilroy closed this Feb 14, 2026

richiemcilroy reopened this Feb 14, 2026

tembo bot reviewed Feb 14, 2026

View reviewed changes

cursor bot reviewed Feb 14, 2026

View reviewed changes

greptile-apps bot reviewed Feb 14, 2026

View reviewed changes

richiemcilroy closed this Feb 14, 2026

-                            if buffered_wait_prefetch_changed {
+                            if seek_rx.has_changed().unwrap_or(false) {
+                                continue;
+                            }
+                            if buffered_wait_prefetch_changed {
+                                trim_prefetch_buffer(&mut prefetch_buffer, frame_number);
+                            }

Conversation

richiemcilroy commented Feb 13, 2026 • edited by greptile-apps bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Overview

Greptile Summary

Confidence Score: 2/5

Important Files Changed

Sequence Diagram

Uh oh!

cursor bot commented Feb 13, 2026

Uh oh!

tembo bot Feb 14, 2026

Choose a reason for hiding this comment

Uh oh!

cursor bot left a comment

Choose a reason for hiding this comment

This PR is being reviewed by Cursor Bugbot

Uh oh!

cursor bot Feb 14, 2026

Choose a reason for hiding this comment

Seek dedupe blocks valid repeat seeks

Uh oh!

cursor bot Feb 14, 2026

Choose a reason for hiding this comment

Repeated same-frame seeks are ignored

Uh oh!

cursor bot Feb 14, 2026

Choose a reason for hiding this comment

Unmount doesn’t cancel pending seek pipeline

Uh oh!

cursor bot left a comment

Choose a reason for hiding this comment

This PR is being reviewed by Cursor Bugbot

Uh oh!

cursor bot Feb 14, 2026

Choose a reason for hiding this comment

Seek dedupe uses stale state

Uh oh!

cursor bot Feb 14, 2026

Choose a reason for hiding this comment

Fragment merge can exhaust memory

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 14, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 14, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 14, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 14, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 14, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 14, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

richiemcilroy commented Feb 13, 2026 •

edited by greptile-apps bot

Loading