v0.4.3

delugemu v0.4.3

Mac fixes for the v0.4.x scaling regressions.

Display: the macOS console window opened tiny because it was launched with
cocoa zoom-to-fit=on, which makes the window resizable; in that mode QEMU
sizes only the aspect ratio of the 640x480 init frame instead of the
front-panel surface. The cocoa path now opens non-resizable, so the window is
sized to the surface (native / backing scale factor, a crisp 1:1 window on
Retina); the View > Zoom-to-fit menu item still toggles it. --skin-scale now
defaults to 'native' (no down-sample) rather than the auto monitor probe,
which could mis-detect and shrink the window enough to make the encoder
rotation triangles unclickable.

Audio: the host playback buffer default is now platform-aware - 30 ms on
macOS/Linux (serviced off the main loop) versus 80 ms on Windows (dsound
pumped from the main loop) - cutting the press-to-sound latency on macOS.

Docs: new README Performance section on the jitter/buffer/latency trade-off
and host CPU throughput limits.

v0.4.2

delugemu v0.4.2

Display: the front-panel window now fits your monitor. Previously the host
window opened at the native panel size (2256x1584) and overflowed smaller
screens, with zoom-to-fit only scaling on later resize. The skin device now
renders the composited panel down-sampled (box-filtered from native) and opens
the console at a screen-fitting size; run.sh detects the primary monitor and
picks the largest scale that fits ~90% of it. New --skin-scale <percent|auto|
native> overrides it; zoom-to-fit still tracks resizes.

Also: add run.sh --icount opt-in (deterministic, artifact-free audio capped to
real time), keep the firmware render-head clamp as the graceful-degradation
default (--tx-render-head), and force zoom-to-fit on the console window.

v0.4.1

delugemu v0.4.1

Fix: audio is now firmware-independent. The SSI TX audio pump bounded its ring copy by a hardcoded firmware symbol address (AudioEngine::i2sTXBufferPos) that was only correct for the in-repo dev firmware build, so the community 1.2.1 release binary the bundle ships produced no audio on any platform. The pump now bounds the copy by the virtual-time DMA play head (CRSA), making it work with any firmware. Also reverted the v0.4.0-era dsound->sdl default change, which was based on a misdiagnosis of this issue.

v0.3.0

DelugEmu v0.3.0

Live-play latency and folder-backed SD card.

Audio / display:
- deluge-skin now skips recompositing unchanged frames, removing the
  periodic full-panel redraw stall that dominated audio latency when
  playing the emulator live from external MIDI.
- SSIF output buffer is runtime-tunable via run.sh --audio-buffer <ms>
  (device property rza1l-ssif.prime-ms); default lowered to 15 ms now
  that the redraw stall is gone.
- Default to the console (front-panel skin) display; the skin PNG is
  bundled in the release and the ssif.in audio error is silenced.

SD card:
- --sd now accepts a directory, snapshotted into a temporary FAT image
  at launch. Directories whose name ends in '_rw' have guest changes
  written back on clean exit. When --sd is omitted, run.sh defaults to
  ./sdcard_rw or ./sdcard.

MIDI:
- New scripts/midi_route.py CoreMIDI router (bundled in releases) wires
  the emulated Deluge to host gear (e.g. a Synthstrom Summit) in both
  directions.

CI attaches a self-contained, relocatable macOS bundle
(DelugEmu-macos-<arch>.tar.gz) to this release.