Open
Conversation
…ions - Batch upstream IRC lines into single IPC messages to reduce overhead - Add fast-path for numeric IRC responses and common replyrouter commands - Direct-stream WHO replies to requesting client, bypassing hooks/replyrouter - Debounce connection state saves to reduce disk I/O by ~95% - Flush dirty connection states on graceful shutdown - Optimize SQLite message queries with covering indexes and pagination - Add priority write path for user commands in sockets layer - Fix setNoDelay to only apply to TCP sockets, not WebSockets - Fix replyrouter array mutation bug causing lost WHO/LIST replies - Fix shared message object mutation in extended-join hook - Preserve nick case when restoring buffer state from database Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…liability - Handle AMQP channel closure with auto-reconnect and retry on send - Auto-connect networks in KiwiIRC when BNC reports them as connected - Create query buffers on outgoing PM/NOTICE for proper lastSeen tracking - Guard lastSeen access in LISTBUFFERS to prevent errors - Write JOIN/TOPIC synchronously in dumpChannels to fix NAMES ordering - Flush buffer after dumpChannels for non-bouncer clients - Don't forward PRIVMSG/NOTICE for channels we've left Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- ConnectionIncoming: dumpChannels, sendNames, writeMsg, lastSeen - ConnectionOutgoing: messageFromUpstream forwarding, nick tracking - ConnectionState: debounced saves, tempSet, getOrAddBuffer, fromObj nick case - Hooks: built-in hooks for multi-prefix, userhost-in-names, WHO fast-path - ReplyRouter: route management, reply forwarding, cleanup - Add shared test mocks and helpers Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Contributor
|
Nice work @CoryChaplin I think it would be better if the database cache and mmap sizes where exposed via the config with defaults set lower to better accommodate vps's with low memory sizes and admins who only want to host a few users. (with a comment on suggestions for bigger networks) As for my question are you sure the priority message system could not result in ordering issues if two commands that are hard to differentiate the results for are sent in quick succession. not sure on a good example but I'm quite sure there is parts of irc-framework that rely on the order of messages to direct responses to the correct cache/request. |
CoryChaplin
force-pushed
the
fix/perf-boost
branch
from
8cc11fb to
0193aaa
Compare
…ssages The batch event handler was registered twice in listenToQueue(), causing every upstream message to be processed and relayed to clients twice. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…jection Wrap the retry sendToQueue call in a try-catch so that if the newly reconnected channel is also closed (race condition), the error is logged and the message dropped instead of propagating as an unhandledRejection.
When the upstream server supports the echo-message capability, outgoing PRIVMSG/NOTICE were being stored twice in the history database: once in clientcommands (immediately on send) and once in upstreamcommands (when the server echoed the message back). This caused the user's own messages to appear twice when reloading history in KiwiIRC. Fix: skip storing in clientcommands when the server supports echo-message, letting upstreamcommands handle storage via the server echo instead.
…toreMessageLoop
Raw exec('BEGIN')/exec('COMMIT') does not update better-sqlite3's db.inTransaction
property, so the retention cleanup code wrongly assumes no transaction is open and
starts its own write transaction concurrently, causing SQLITE_BUSY.
Replace with db.transaction() which properly tracks transaction state, and make
dataId() synchronous (it had no actual async operations) so it can run inside the
synchronous transaction wrapper. Switch the recursive storeMessageLoop call to
setImmediate to prevent stack overflow on large queues.
Expose cache_size and mmap_size as config options with conservative defaults (2MB/disabled) for small VPS hosts. Remove MODE from priority commands to prevent potential response-routing issues with its dual action/query semantics.
CoryChaplin
force-pushed
the
fix/perf-boost
branch
from
8e4bb5c to
878c1c8
Compare
Contributor
Author
|
@ItsOnlyBinary Here's an updated version vs your comments, and some production issues fixed. |
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.
Quite a big evolution, but it produces a dramatic improvement to the bnc's performance on large networks.
It has been running on my instance for a while now with no issues.
Here's a view on the IOs improvements.