CHANGELOG.md

Changelog

All notable changes to JamWide will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

1.1 Beta Series

The 1.1 line is a complete JUCE rewrite (1.0 was CLAP/ImGui). Per-beta release notes for 1.1-beta.1 through 1.1-beta.20 are on the GitHub Releases page and are not duplicated here. Entries are tracked below starting with 1.1-beta.20.1.

[1.1.0-beta.20.5] - 2026-05-05 — DIAGNOSTIC BUILD (pool + profiling)

Recommended diagnostic build for the multi-peer CPU-spike regression test. Supersedes .20.3 and .20.4. The .20.4 ABTEST 1 (ring 256→32) was redesigned because reverting the size re-introduced the original cutoff bug at typical bitrates, which mixed in a different audio-drop mechanism and confused the test signal. This build keeps the 256-chunk capacity (cutoff fix preserved) and instead pools DecodeMediaBuffer instances so per-interval allocation is eliminated entirely after warm-up.

Changed (vs .20.4)

• ABTEST 1 redesigned: instead of reverting SpscRing<DecodeChunk, N> from 256 to 32 chunks, we keep N=256 and introduce DecodeMediaBufferPool. RemoteDownload::Open now acquires from the pool (new only on cold path, when the free list is empty); DecodeMediaBuffer::Release() returns to the pool instead of delete this. DecodeMediaBuffer::ResetForReuse() drains any leftover SPSC chunks and zeroes per-instance counters before recycling. Pool capped at 64 instances (~64 MB resident; covers 64-peer rooms).
• Side effect of the pivot: cutoff bug from .20.3/.20.4 is no longer re-introduced. The pool preserves 0e9cbae's cutoff fix.

Same as .20.4 (still active)

• ABTEST 2: broadcastBeatHeartbeat stubbed (heartbeat baseline-CPU test). Side effect: video companion sync indicator stops updating.
• Profiling: client->Run(), RemoteDownload::Open acquisition, and JamWideJuceEditor::timerCallback instrumented with count/total_ns/max_ns counters. Surfaced in /rcmstats and DBG button output.

What the profiling now reveals

With the pool, the decbuf_alloc avg_ns metric should trend toward drain-and-reset cost (sub-µs) not mach_vm_allocate cost (often hundreds of µs). If the spike still occurs despite this, the allocation hypothesis is falsified and we'd look elsewhere. If the spike resolves, this build is essentially the production fix candidate.

Tester protocol

Same as .20.4: DBG at start, jam 30 min with 4+ peers, DBG at end, diff the two log files.

Key signals to look at:

• decbuf_alloc avg_ns delta across the session: pool hot path should be < 10 µs (was ~hundreds of µs on .20.2 / .20.3 cold path)
• client_run max_ns: should drop significantly if allocation churn was the spike source
• Audible glitches and remote-peer reports: should be absent if hypothesis confirmed

v1.1-beta.20.4 status

Tag v1.1-beta.20.4 exists on the remote but no GitHub release was created — its CI was cancelled before the release job ran, in favor of this redesign. The commit (f36a4a6) remains in main's history as the prior diagnostic checkpoint.

[1.1.0-beta.20.4] - 2026-05-04 — DIAGNOSTIC BUILD (with profiling)

Same diagnostic intent as .20.3 (CPU-spike A/B test) plus profiling counters that the log file can carry. Recommended diagnostic build — supersedes .20.3 for testing.

Added (profiling)

• client->Run() cycle time — count, total, max ns (run thread)
• RemoteDownload::Open new DecodeMediaBuffer allocation time — count, total, max ns (run thread)
• JamWideJuceEditor::timerCallback end-to-end time — count, total, max ns (message thread)
• All counters surfaced via /rcmstats chat command and DBG button — appended to existing diagnostic report under --- profiling (cumulative; compare deltas) ---

Same as .20.3 (still active)

• ABTEST 1: SPSC ring 256 → 32 (allocation churn test)
• ABTEST 2: broadcastBeatHeartbeat stubbed (message-thread baseline test)

How to use the profiling

1. Connect to a populated server, wait for steady state (4+ peers)
2. Run /rcmstats — note the three profiling rows; this is your start snapshot
3. Jam for 30 minutes
4. Run /rcmstats again — note the three profiling rows; this is your end snapshot
5. Compute deltas (end − start). Key metrics:
   • decbuf_alloc avg_ns — if it's > 100 µs (100,000 ns) on production sizing, libmalloc is going to mach_vm_allocate. Confirms hypothesis.
   • decbuf_alloc max_ns — peak single-allocation time. > 1 ms means a single alloc was a real spike.
   • client_run max_ns — peak run-thread cycle. > 5 ms means audio data was withheld from the encoder upload.
   • timer_cb avg_ns — message-thread baseline cost. Compare ABTEST-2-on vs ABTEST-2-off builds.

Tester protocol

Side-by-side test: v1.1-beta.20.2 (no fixes, no profiling) vs v1.1-beta.20.4 (both fixes + profiling). Both should run against the same room population for the same duration. Note: only .20.4 will show profiling — but the symptom-presence/absence comparison itself is the primary signal.

[1.1.0-beta.20.3] - 2026-05-04 — DIAGNOSTIC BUILD

⚠ This is not a regular beta. It is a diagnostic build for the multi-peer CPU-spike investigation tracked in .planning/debug/cpu-spikes-beta12-regression.md. Two suspect code paths are temporarily neutralized to confirm whether they are the regression source. Do not use this build for normal jamming — at high bitrates the original audio-cutoff bug that beta.20.1 fixed is intentionally re-introduced. Use beta.20.2 for normal use.

Diagnostic changes (NOT shipping fixes)

• ABTEST 1: DecodeMediaBuffer's SPSC ring temporarily reverted from 256 → 32 chunks. Confirms whether per-peer-per-interval 1 MB heap allocation churn (introduced by beta.20.1's ring bump) is the cause of audible CPU spikes that grow with peer count. Side effect: original interval-overflow cutoff bug returns at high bitrates (≥192 kbps stereo on 12 s intervals); decbuf_drops will climb in /rcmstats.
• ABTEST 2: broadcastBeatHeartbeat call in JamWideJuceEditor::timerCallback is stubbed. Confirms whether ~1.5 Hz JSON build + WebSocket send on the message thread is the source of the baseline CPU bump that started showing in beta.12. Side effect: video companion sync indicator stops updating; no other functional impact.

Notes for testers

• A/B compare baseline (beta.20.2) against this build under the same load (4+ peers, 30+ min session)
• Watch for: (a) audible-glitch cadence change, (b) Activity Monitor CPU% baseline, (c) decbuf_drops counter via /rcmstats
• Report findings to inform the proper fix (pool DecodeMediaBuffer + rate-limit/move heartbeat off message thread)

[1.1.0-beta.20.2] - 2026-05-03

Fixed

• Chat: Scrolling back through history is now sticky — new messages no longer yank the viewport down to a hidden caret
• Windows: Build unbroken (std::strftime collision with WDL macro from beta.20.1's diagnostic-counter additions)

[1.1.0-beta.20.1] - 2026-05-03

Added

• Diagnostics: DBG button in UI dumps counters to a log file for bug reports
• Diagnostics: /rcmstats chat command types remote-channel-mirror counters into chat for live triage
• Diagnostics: Local-channel mirror snapshot and IsNetConnected accessor for UI use without piercing audio-thread state

Changed

• Audio: DecodeMediaBuffer's SPSC ring grew from 32 to 256 chunks for more head-room under codec/network timing variance

Fixed

• UI: VU meter scale now matches the fader scale (was using a different dB-to-pixel mapping)
• UI: Chat auto-scrolls on new messages
• UI: Local strip TX button visual stays in sync with localTransmit[0]
• UI: Strip-keyed remote params reset cleanly on plugin load (no more ghost values from prior sessions)
• Audio: dump_samples skip-debt accumulation removed; codec underruns recover cleanly without biasing subsequent intervals
• Network: PeerChannelInfoUpdate now wired through the SPSC mirror, fixing stale UI when a remote user renames a channel mid-session

[1.0.0] - 2026-01-14

🎉 First Stable Release

JamWide 1.0 is the first stable release of the NINJAM client plugin.

Highlights

• Full NINJAM protocol support (connect, transmit, receive audio)
• Cross-platform: macOS (Intel + Apple Silicon) and Windows
• Plugin formats: CLAP, VST3, Audio Unit v2
• Tested in: Ableton Live, REAPER, Bitwig Studio, Logic Pro, GarageBand

All Features

• Server browser with live user lists (autosong.ninjam.com)
• Real-time chat with message history and timestamps
• Visual timing guide for beat alignment
• Per-channel volume, pan, mute, and solo controls
• VU meters for all channels
• BPM/BPI voting via chat commands
• Anonymous login support
• 256 kbps default audio quality (OGG/Vorbis)
• State persistence (save/load with DAW projects)
• Parameter automation (master volume/mute, metronome volume/mute)

Changed

• Removed unused REAPER integration code (hwnd_info API)
• Cleaned up final NINJAM→JamWide naming references

Fixed

• All known issues from beta testing resolved

[0.133] - 2026-01-14

Fixed

• All Platforms: Solo channel no longer crashes (mutex deadlock fix)
• Windows: Fixed keyboard duplication in text fields (keys no longer repeat)

[0.131] - 2026-01-13

Added

• macOS/REAPER: Show hint to enable "Send all keyboard input to plug-in" for full keyboard support
• Chat: Input field now keeps focus after sending a message

Fixed

• macOS: Simplified keyboard handling - removed experimental swizzle/monitor code
• Chat: Send button positioning improved (no longer clips at edge)

[0.119] - 2026-01-12

Fixed

• Windows: Message hook implementation prevents DAW accelerators from triggering during text input
• Windows: Spacebar no longer triggers DAW transport when typing in text fields
• Windows: Caps Lock now works correctly in text fields (Bitwig/REAPER)

Changed

• System Requirements: Windows 10 or later now required (Windows 7/8 no longer supported)

[0.117] - 2026-01-12

Added

• Windows: Dummy EDIT control for proper keyboard focus signaling to DAW
• Windows: IME support for Japanese/Chinese/Korean keyboard input
• Windows: Focus event forwarding (WM_SETFOCUS/WM_KILLFOCUS)

Fixed

• Windows: Keyboard input now works correctly in text fields
• Windows: Added null guard for orig_edit_proc_ to prevent crashes if subclassing fails

[0.116] - 2026-01-12

Changed

• Windows: Initial keyboard focus implementation with dummy EDIT control

[0.108] - 2026-01-12

Fixed

• UI: Transmit toggle now visible (layout fix)

[0.107] - 2026-01-11

Fixed

• License dialog now responds to single click instead of requiring double-click

[0.106] - 2026-01-11

Changed

• Default audio quality increased to 256 kbps (highest quality)

[0.105] - 2026-01-11

Added

• Server browser now displays usernames from autosong.ninjam.com

[0.104] - 2026-01-11

Changed

• Audio Unit window size fixed at 800x1200 for Logic Pro/GarageBand compatibility
• Implemented setFrameSize handler for AU

[0.90] - 2026-01-10

Added

• Visual timing guide with beat grid and transient dots
• Chat room with message history, timestamps, and input field
• ImGui ID collision fixes throughout UI
• Release automation script (release.sh)

Fixed

• Anonymous login now auto-prefixes "anonymous:" for public servers

[0.1.0] - 2026-01-07

Added

• Initial CLAP plugin implementation
• NINJAM client core ported from ReaNINJAM
• Cross-platform GUI (ImGui + Metal/D3D11)
• Server browser with live server list
• Connection management (connect/disconnect)
• Local channel controls (volume/pan/mute/transmit)
• Remote user channels with per-channel controls
• Master and metronome controls
• VU meters for all channels
• License agreement dialog
• State persistence (JSON save/load)
• Parameter automation support (4 params)
• Multi-instance support
• Thread-safe architecture with command queues
• Windows build system (Visual Studio 2022+, PowerShell)
• macOS build system (Xcode, bash)
• GitHub Actions CI/CD for automated builds

Supported

• Plugin formats: CLAP, VST3, Audio Unit v2
• Platforms: macOS 10.15+, Windows 10+
• DAWs: Logic Pro, GarageBand, Bitwig Studio, REAPER, and more