release #5541: iOS PWA streaming freeze + SSE reconnect reliability - #5586
Merged
Merged
Conversation
Two mobile reliability fixes on the crown-jewel chat streaming path: (1) content-visibility:auto on off-screen .msg-row under @media(pointer:coarse) so WKWebView skips layout/paint for off-screen rows during streaming (kills the long-chat freeze). The LIVE turn is kept content-visibility:visible via the STABLE #liveAssistantTurn id (covers ALL render modes — Compact Worklog, Transparent Stream, restored-live — not just the Transparent-Stream path that stamps data-live-assistant-turn), so a normal live turn on touch never blanks mid-stream and its height keeps growing so the new-message cue still fires. contain-intrinsic-size:auto 1px preserves flick-scroll momentum. Scoped to touch — desktop find-in-page untouched. Deliberately does NOT set overflow-anchor:none (inert on iOS WebKit; re-opens #4856/#5338). (2) SSE reconnect ladder extended 4->6 steps + a last-ditch _restoreSettledSession full-session poll (8s watchdog) after retries exhaust, so a response completed during an iOS Tailscale/VPN reconnect is recovered without an error banner. Co-authored-by: luperrypf <luperrypf@users.noreply.github.com>
Contributor
|
| Filename | Overview |
|---|---|
| static/messages.js | Extends SSE retry ladder from 4 to 6 steps and adds a last-ditch _restoreSettledSession with an 8s watchdog; the watchdog covers only the second call, leaving the pre-existing probe-loop call unguarded |
| static/style.css | Adds content-visibility:auto + contain-intrinsic-size:auto 1px + contain:layout style to .msg-row under @media (pointer:coarse), with a correctly-targeted #liveAssistantTurn ID selector to keep the live turn fully rendered on all paths |
| tests/test_sse_error_multi_probe_reconnect.py | Correctly updated static assertions to reflect the expanded 6-step delay array and new last-ditch reconnect structure; brace-matching approach is robust against future growth |
| CHANGELOG.md | Release-managed CHANGELOG entry added for both the streaming-freeze fix and the SSE reconnect improvements; no issues |
Sequence Diagram
%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
participant ES as EventSource
participant PR as _probeReconnect
participant SS as /api/stream/status
participant RS as _restoreSettledSession
participant UI as Composer UI
ES->>PR: onerror fires
PR->>UI: "Reconnecting… (1/6)"
loop For each of 6 retry delays
PR->>SS: GET /api/chat/stream/status
alt stream active
SS-->>PR: "{active:true}"
PR->>ES: reconnect EventSource
else replay available
SS-->>PR: "{replay_available:true}"
PR->>ES: reconnect with replay
else still not settled
SS-->>PR: "{}"
PR->>RS: _restoreSettledSession() [no watchdog]
Note over PR,RS: If RS hangs here on final probe, watchdog never starts
alt more retries remain
PR->>UI: "Reconnecting… (N/6)"
PR->>PR: schedule next probe
end
end
end
Note over PR: All 6 probes exhausted
PR->>UI: "Restoring session…"
PR->>RS: _restoreSettledSession() [8s watchdog]
alt RS resolves in time
RS-->>PR: true/false
PR->>PR: clearTimeout(watchdog)
else watchdog fires first (8s)
PR->>UI: _handleStreamError
end
%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
sequenceDiagram
participant ES as EventSource
participant PR as _probeReconnect
participant SS as /api/stream/status
participant RS as _restoreSettledSession
participant UI as Composer UI
ES->>PR: onerror fires
PR->>UI: "Reconnecting… (1/6)"
loop For each of 6 retry delays
PR->>SS: GET /api/chat/stream/status
alt stream active
SS-->>PR: "{active:true}"
PR->>ES: reconnect EventSource
else replay available
SS-->>PR: "{replay_available:true}"
PR->>ES: reconnect with replay
else still not settled
SS-->>PR: "{}"
PR->>RS: _restoreSettledSession() [no watchdog]
Note over PR,RS: If RS hangs here on final probe, watchdog never starts
alt more retries remain
PR->>UI: "Reconnecting… (N/6)"
PR->>PR: schedule next probe
end
end
end
Note over PR: All 6 probes exhausted
PR->>UI: "Restoring session…"
PR->>RS: _restoreSettledSession() [8s watchdog]
alt RS resolves in time
RS-->>PR: true/false
PR->>PR: clearTimeout(watchdog)
else watchdog fires first (8s)
PR->>UI: _handleStreamError
end
Reviews (1): Last reviewed commit: "fix(#5541): iOS PWA streaming freeze + S..." | Re-trigger Greptile
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
Comment on lines
+5797
to
5835
| // Last-ditch: the stream may have finished while we were retrying. | ||
| // _restoreSettledSession polls the full session API (not just stream | ||
| // status) and can recover a completed response without an error banner. | ||
| // This is especially important on iOS where Tailscale reconnects can | ||
| // take longer than the retry window. | ||
| setComposerStatus('Restoring session…'); | ||
| let _restoreTimedOut=false; | ||
| const _restoreTimer=setTimeout(()=>{ | ||
| // If _restoreSettledSession hangs (flaky Tailscale), don't leave | ||
| // the UI stuck on "Restoring session…" forever. Fall through to | ||
| // _handleStreamError after 8s. | ||
| _restoreTimedOut=true; | ||
| if(!_terminalStateReached&&!_streamFinalized){ | ||
| if(_deferStreamErrorIfOffline()) return; | ||
| if(_deferStreamErrorIfPageHidden(source)) return; | ||
| _flushReasoningToAnchor(); | ||
| _scheduleAnchorRegistryCleanup(120000); | ||
| _handleStreamError(source); | ||
| } | ||
| },8000); | ||
| try{ | ||
| if(await _restoreSettledSession(source, {preserveVisibleOnShorterTerminalSnapshot:true})){ | ||
| if(_restoreTimedOut) return; // timer already fired _handleStreamError | ||
| clearTimeout(_restoreTimer); | ||
| return; | ||
| } | ||
| }catch(_){ | ||
| // _restoreSettledSession threw. If the timer already fired, | ||
| // _handleStreamError was called there; we return below. | ||
| // Otherwise the code below cancels the timer and calls it directly. | ||
| } | ||
| if(_restoreTimedOut) return; // timer already fired _handleStreamError | ||
| clearTimeout(_restoreTimer); | ||
| if(_terminalStateReached||_streamFinalized) return; | ||
| if(_deferStreamErrorIfOffline()) return; | ||
| if(_deferStreamErrorIfPageHidden(source)) return; | ||
| _flushReasoningToAnchor(); | ||
| _scheduleAnchorRegistryCleanup(120000); | ||
| _handleStreamError(source); |
Contributor
There was a problem hiding this comment.
_restoreSettledSession twice, but only the second call is guarded
On the last probe (attempt=5, where nextDelay is undefined), the code first falls through to the existing _restoreSettledSession call at line 5788, which has no timeout. Only if that returns false does execution reach the new last-ditch block at line 5804 where the 8s _restoreTimer starts. If the line-5788 call hangs on a flaky Tailscale link — the exact scenario the watchdog is designed to handle — the timer never gets created, leaving the UI stuck on "Reconnecting… (6/6)" indefinitely instead of the intended 8s cap. The watchdog is well-designed for the second call but doesn't cover the prior call in the same probe.
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.
Release: #5541 — iOS PWA streaming freeze + SSE reconnect reliability
Ships @luperrypf's mobile streaming reliability fix, rebuilt on current
master, with a gate-caught defect fixed and the full gate green. Nathan approved shipping on the green gate.The bug (#5541)
On iOS/Android PWA, a long streaming conversation could freeze — WKWebView re-lays-out the whole document on every
scrollHeightread as the response grows. Separately, a dropped SSE connection (iOS Tailscale/VPN reconnect) could exhaust the retry ladder and leave a permanently dead stream needing a manual refresh.The fix
@media (pointer: coarse)): off-screen.msg-rows getcontent-visibility:auto+contain-intrinsic-size:auto 1px+contain:layout style, so WKWebView skips their layout/paint during streaming. The live turn is kept fully rendered via the stable#liveAssistantTurnid (covers every render path — Compact Worklog, Transparent Stream, restored-live) plus thedata-live-assistant*forms, so it never blanks mid-stream and its height keeps growing so the new-message cue still fires. Scoped to touch — desktop rendering is pixel-identical. Deliberately does not setoverflow-anchor:none(inert on iOS WebKit; re-opens the bug: WebUI is unusable on Android - transcript jumps to top on every interaction (regression in v0.51.576) #4856/fix(chat): suppress browser overflow-anchor during JS scroll-anchor realign (mobile scroll jump-back) #5338 Android jump-to-top regression).[1500,3000,5000,8000]→+[12000,20000], then a last-ditch_restoreSettledSessionfull-session poll (8s-watchdog-guarded, with a "Restoring session…" affordance) recovers a response that completed while the connection was flapping — without an error banner.Gate-caught defect (fixed this round)
The first gate found the live-turn force-visible selector only covered
data-live-assistant-turn(stamped only by the Transparent Stream path) — so a normal live turn (#liveAssistantTurnwithout the attribute) on touch could becontent-visibility-skipped and blank out mid-stream. Fixed by adding#liveAssistantTurn(the stable id set on all render paths) to the force-visible selector. Verified live in-browser.Gate (green)
#liveAssistantTurnforce-visible confirmed live, desktopcontent-visibilityunaffected (fine-pointer), zero JS console errors.Attribution: original author @luperrypf (
Co-authored-bytrailer preserved).Closes #5541