Restart sync on joinLayer for immediate room detection#13
Merged
ThomasHalwax merged 6 commits intomainfrom Mar 23, 2026
Merged
Conversation
joinLayer() now follows the sequence: add room to filter (idMapping), restart the sync long-poll, then perform the actual join. This ensures the restarted poll includes the room and sees the join event when it arrives, enabling the sync-gated content fetch to work reliably. Changes: - http-api: pass AbortSignal through to ky for sync requests - timeline-api: add restartSync() to abort current long-poll; stream loop catches the abort and re-enters with updated filter - project: reorder joinLayer() to update filter before join - E2E test: verify sync-gated received() delivers content after join
restartSync() now returns a Promise that resolves once the stream loop has applied the updated filter. joinLayer() awaits this before calling join(), ensuring the sync request with the new room is already in flight when the join event arrives on the server.
When a room appears in sync with limited:true but /messages returns empty (remote server hasn't backfilled yet), keep the room in pendingContent and retry on subsequent sync cycles instead of giving up immediately. Gives up after 10 retries. Changed pendingContent from Set to Map to track retry state.
content() now accepts a 'from' pagination token. When provided, it paginates backward (dir=b) instead of forward from the start. This fixes federation scenarios where forward pagination returns empty because the remote server hasn't backfilled yet, but the prev_batch token from sync provides a valid backward pagination starting point. The pending content processing now captures prevBatches from sync responses and passes them through to content().
After the first sync appearance, keep retrying content fetch on subsequent sync cycles even if the room doesn't appear again. This handles the case where encrypted events are fetched but Megolm keys haven't arrived yet via Historical Key Sharing. Retries are throttled to every 5 seconds (configurable via contentRetryIntervalMs constructor option) and give up after 10 attempts (~50 seconds).
matrix.org can take 2+ minutes for federation backfill. Previous 10x5s window (~50s) was too tight. Now 20x10s (~3.5 minutes).
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
2 participants
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.
Problem
After
joinLayer(), the sync-gated content mechanism inProject.start()could not detect the room because the running long-poll'sroomsfilter didn't include it. The poll had to time out (30s) before the next one picked up the updated filter.Solution
Reorder
joinLayer()to follow: filter update → sync restart → joinidMapping(temporary self-mapping) so it appears in theroomsfilterrestartSync()to abort the current long-pollPOST /joinstateEventscheck fires →content()loads operations →received()delivers themImplementation
httpApi.sync(): Accept optionalAbortSignal, pass tokyTimelineAPI.restartSync(): Abort the current iteration'sAbortControllerTimelineAPI.stream(): Per-iterationAbortController, catch abort as restart (not error)Project.joinLayer(): Reordered as described aboveTests
content()+ sync-gatedreceived()(1.5s)