1.3.0

This release makes Mosaic’s sequencing substantially more reliable under load.

• Fixed timing and lifecycle bugs affecting MIDI clock and transport, parameter locks and slides, recording, scales, chords, arpeggios, song playback, project loading, and stuck notes.
• Clock timing is now continuously measured against musical deadlines. In 16-channel hardware stress tests, 99% of events arrived within 3.9 ms, with final phase error below 2.5 ms - so timing did not progressively drift.
• Hardware qualification covers dense playback, parameter locks and slides at 130 BPM, plus an extreme 200 BPM workload producing approximately 1,080 MIDI messages per second. The release gates require p99 timing below 10 ms, no event beyond 50 ms, final phase error below 20 ms, and step jitter below 10 ms.
• The lock, slide, and 200 BPM extreme workloads passed every measured window. Dense playback passed five of six; one isolated 11.1 ms jitter spike exceeded the strict 10 ms gate and remains recorded rather than hidden.
• monome-emulator now runs the real Mosaic application in a reproducible norns environment, driving grid, keys, encoders, MIDI, clocks, audio and connected-device paths. It supports exact controlled-time tests, real-time behaviour tests, performance profiling, and CI coverage without requiring manual hardware testing.

See PR #94 for full details.

1.2.12

• Fixed regression where norns devices wouldn't initiate due to lack of preexisting state

1.2.11

• Fixed issue where params dont display correctly in norns param menu

1.2.10

• Fixed regression that was causing manual song mode changes to not refresh pattern state

1.2.9

• Further song mode parameter handling fixes

1.2.8

• Song mode now respects repeat settings when considering which first step to process.

1.2.7

• Fixed the reverse case, where param locks on the first step of a pattern were not being applied properly outside of song mode.

1.2.6

• Fixed issue where param locks on the first step of a pattern were not being applied properly in song mode.

1.2.5

Fixed regression that led to midi input not using the correct global scale number

1.2.4

• Fixed inconsistent behaviours with chord masks being quantised to the scale degree and rotations depending on different settings.