taud bugfix

This commit is contained in:
minjaesong
2026-05-09 21:17:49 +09:00
parent f69108c40d
commit ddeab1c782
3 changed files with 44 additions and 7 deletions

View File

@@ -182,6 +182,26 @@ def parse_mod(data: bytes):
inst = (b0 & 0xF0) | ((b2 >> 4) & 0x0F)
effect = b2 & 0x0F
arg = b3
# MT-style PT-strict cell rewrites (LoaderMOD.cpp:354-365):
# PT does not recall arg for portamento up/down (1xx, 2xx) or
# volume slide (Axx); the literal arg is read every tick. The
# vol-slide nibbles in 5xx/6xx likewise take literal args, with
# the recalled state living in the porta/vibrato side. So a
# zero-arg cell decays to a no-slide variant: 1/2/A drop to
# no-op, 5 collapses to bare tone-porta (3), 6 to bare vibrato
# (4). Without this, resolve_pt_recalls would back-fill these
# zero args from the cohort memory and produce a continuous
# slide where PT plays a single-row swell (canonical bug:
# GSLINGER ord 0x03 ch1 — `5 01` on r30/r38 with `5 00` on the
# rest, was fading 24→0 in 5 rows instead of stair-stepping
# 24→14 across 16 rows).
if arg == 0:
if effect in (0x1, 0x2, 0xA):
effect = 0x0
elif effect == 0x5:
effect = 0x3
elif effect == 0x6:
effect = 0x4
cell = grid[ch][r]
cell.period = period
cell.inst = inst

View File

@@ -2394,8 +2394,8 @@ TODO:
when no V column is present. Engine + all four `*2taud` converters
updated; legacy `.taud` files (byte 196 == 0) fall back to the
previous "row volume default = 63" behaviour.
[ ] Physical Presence order 0x1F chn 2: note cuts unexpectedly fast?
GSLINGER order 0x03 chn 1: L 0100 fades unexpectedly fast?
[x] physical_presence order 0x1F chn 2: note cuts unexpectedly fast — engine fix
[x] GSLINGER order 0x03 chn 1: L 0100 fades unexpectedly fast? — converter fix
TODO - list of demo songs that MUST ship with Microtone:
* 4THSYM (rename to Fourth Symmetriad) — excellent piece for demonstrating NNAs and filter envelopes

View File

@@ -2145,11 +2145,28 @@ class AudioAdapter(val vm: VM) : PeriBase(VM.PERITYPE_SOUND) {
// mirroring G's behaviour — the L command continues the porta started by an earlier G.
val toneG = (row.effect == EffectOp.OP_G || row.effect == EffectOp.OP_L)
when (row.note) {
// No note but an instrument byte is present: latch the instrument so
// the *next* note-only trigger picks up the right sample. Trackers
// call this an "instrument-only retrigger"; in MOD/S3M/IT the sample
// keeps playing, but the channel's instrument reference advances.
0xFFFF -> { if (row.instrment != 0) voice.instrumentId = row.instrment }
// No note but an instrument byte is present: latch the instrument and
// re-seed the channel volume from the new sample's Default Note Volume.
// PT, FT2, IT and Schism all do this — pt2_replayer.c:1086 writes
// ch->n_volume = s->volume on every sample-byte row regardless of note;
// ft2_replayer.c:1431-1434 calls resetVolumes(ch) when (note==0 && inst>0);
// schism csf_instrument_change writes chan->volume = psmp->volume whenever
// inst_column is set. Without this, a MOD pattern that runs continuous
// volume slides while re-asserting the sample byte each row (e.g.
// physical_presence ord 0x1F ch2: every row carries `... 1E A0F/A09/A02`)
// silences after the first row because the slide saturates at 0 and there's
// nothing to lift the volume back up before the next slide starts.
0xFFFF -> {
if (row.instrment != 0) {
voice.instrumentId = row.instrment
val seedVol = rowVolumeFromDefault(instruments[voice.instrumentId])
voice.channelVolume = seedVol
voice.rowVolume = seedVol
voice.keyOff = false
voice.noteFading = false
voice.fadeoutVolume = 1.0
}
}
// Key-off: release sustain; envelope walks past the sustain point and the fadeout
// begins (foreground-voice fade path at line ~2380). The voice deactivates when
// fadeoutVolume reaches 0, or immediately if FT2-mode fadeStep == 0. Setting