The reopen said "the pool fills and never returns a source". The 30s census in
the SAME field logs disproves it: free returns to ~227-230 between bursts and
reuses climbs ~20/s all session. free=0 on the failure line is true by
DEFINITION at the instant of a failed acquire -- the third instance of the
counter-sampling trap (live=256-vs-6, then hsparm greps, now this), read off the
alarm line instead of the trend line.
WHAT THE LOGS ACTUALLY SHOW
* The pool cycles; release-on-stop exists and works (the engine steal loop).
* During firefights CONCURRENT demand exceeds 240 and the priority steal loop
services each new sound by killing an old one -- continuously through
combat (failures spread across every decile of every combat session).
* Idle standing demand is ~13 sources. My "each component holds its SourceSet
to entity teardown, ~20-25 per mech" narrative was wrong.
* The raw ACQUIRE FAILED line count (6.8k-19k per log) is NOISE: the steal
loop retries after every failed attempt, so lines accumulate per EVENT and
most events still play via a steal. True drops were never counted.
CHANGES
1. Census now carries steals= and drops= (drops = the steal loop ran dry and
the sound NEVER played) plus a per-class drop histogram
("[audio] dropped by class: {class 1005 x4v: N} ...") -- all ungated, so
the next field logs are decisive instead of suggestive.
2. The ACQUIRE FAILED print is rate-limited to 1/30s and now names the
requesting class + voice count. 19k-line log spam distorted this triage.
3. BT_AUDIO_SOURCES=<n> now raises the POOL cap too (it previously raised the
AL context budget while the pool stayed at 240, making the field
experiment impossible to run).
MEASURED (9 mechs, missile autofire, 150s)
* cap 240: peak 130 sources, 0 fails -- demand tracks SHOOTER count, not mech
count; one shooter cannot saturate. A 6-shooter lobby pins 240.
* cap 48 (BT_AUDIO_SOURCES=64): saturation reproduced -- census
steals=223 drops=455, histogram names the classes.
* Dominant field requester (requested=4) = class 1005 Static3DPatchSource:
world-placed effect sounds, i.e. EXPLOSIONS. 1001 DirectPatchSource x1v
dominates drops at low cap; 1002 Dynamic3DPatchSource x3v present.
CONSEQUENCE FOR THE FIX ORDER: #84's stale-aim double detonation duplicates
exactly the saturating class on observer nodes. Fix#84 FIRST, then re-read the
field census; only if it still saturates does the budget experiment
(BT_AUDIO_SOURCES with frame time measured) become the play.
KB: the wrong night-9 entry in open-questions.md replaced with the corrected
diagnosis; gotcha candidate noted -- an alarm-line counter is not a trend.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Found while checking whether a player had reported this before: the code already
carried a prior analysis of this exact error which deliberately decided NOT to
act on it, concluding the 2026-07-23 atomic-delete fix (a999e5c) had settled it
with "no field complaints since", and that the remaining exhaustion was
transient and likely sub-perceptual.
That claim did not hold, and the note was still sitting there to mislead the next
reader. All five 4.11.674 player logs are saturated: 3031/4657/5245/6275/6571
failures, first one ~10% into a match, still failing at 97%.
The prior note's core reasoning was about raising the source BUDGET -- more
voices mixing = more CPU during heavy combat -- and it is sound; BT_AUDIO_SOURCES
stays opt-in and unset. But it does not apply to pooling, which changes no
budget: idle pooled sources are stopped and detached and cost nothing to mix.
The note asks whoever touches this to measure frame time, so I did, same bench
both ways:
pre-fix 8.672 ms weighted avg over 9390 frames
post-fix 7.794 ms weighted avg over 10047 frames
Removing ~20k alGenSources/alDeleteSources driver calls per match is a net CPU
WIN, not a cost. Single run each, so indicative rather than definitive, but it
points the opposite way from the concern.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every 4.11.674 player log is saturated with acquisition failures -- 3031, 4657,
5245, 6275, 6571 across five machines. In Lynx's largest session the first
failure lands 9.3% in and they continue to 96.8%: once it starts it never
recovers for the rest of the match.
CAUSE: RequestAudioChannels called alGenSources() per sound event and
ReleaseSourceSet called alDeleteSources() on release -- create and destroy per
sound. OpenAL sources are a scarce driver resource (OpenAL Soft caps a context
at 256) and a combat burst churned straight through the ceiling.
The two counters looked contradictory and were the tell: ACQUIRE FAILED always
printed live=256 while the 30s census printed live=6. Same global, sampled at
different moments -- sources spike to the cap during a burst and drain back
between them. Churn, not a steady leak.
FIX: generate sources once, up to a cap, and recycle them through a free list.
Release scrubs and parks instead of deleting. Steady-state play performs no AL
allocation at all.
The scrub is load-bearing, not hygiene: a recycled source carries whatever the
previous owner set, and the engine sets AL_LOOPING per sound
(L4AUDLVL.cpp:327). Hand a looping source to a one-shot and it plays forever --
which is the "sound stuck looping" family (#51, #5). Every reusable property is
reset at the single point where a source changes owner.
VERIFIED (scratchpad/night8/audiopool.sh + the two-node mp_burst.sh):
solo, sustained fire : 0 failures, pooled 152, reuses 21998
two nodes, 4 min : 0 failures on both, pooled 148/149, reuses ~7000 each
⚠ HONEST LIMIT: the bench does NOT reproduce the field failure -- the PRE-fix
binary also scores 0 on it (peak 49 live), because solo/2-node combat is not
dense enough to reach 256. So this verifies the pool works and allocates
nothing in steady state; it does NOT by itself prove the field failures are
gone. The five field logs remain the "before".
Peak demand is set by how many audio COMPONENTS are alive (each reserves a
SourceSet of up to 25 voices and holds them), not by audible sounds: measured
high-water 138 solo, 149 two-node. That scales with player count, so the cap is
set near the driver ceiling (240) and the pool now logs its high-water mark once
per 25-source band -- so the next playtest sizes this from field data instead of
a guess. Growth also self-limits: if a driver offers fewer sources than the cap,
alGenSources simply fails, growth stops, and the pool recycles what it has.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
User + old-timers were right; the earlier "no voice exists" verdict was wrong.
The authored mech audio has state watchers on Entity.SimulationState==3/4
(the binary's one-cell mech+0x40) that start sequence notes 29,16,40 -- two
klaxon hits then Warnings01 zone 8 (key 40-41) = the "reverse disabled"
voice line the testers remember. The port's cell split (graphicAlarm vs
engine simulationState, gotcha #23) meant the trigger value never arrived.
- mechdmg: mirror gimp 3/4 into SetSimulationState at the leg-half crossing
(guarded: never stomps disabled/fall/dead states). Voice verified playing
end-to-end on the bench: SetupPatch bank2 patch113 note=40 ->
Warnings01_z7.wav.
- mech.cpp: BT_GIMP_SAFE_BASE_READ on all seven Simulation::ReadUpdateRecord
sites -- gimp is monotonic per life; a record captured pre-gimp must not
stomp the cell (the loopback otherwise perpetuates the stale value and
restarts the warning on every damage event).
- diagnostics (all env-gated): [statefire]/[startreq]/[animind]/[gimp-sim]/
[simstomp]/[indstomp] + StateIndicator::DebugAudioWatcherCount + raised
spatial-log caps. These traced the whole chain and PROVED no SetState
path stomps the cell.
OPEN (follow-up): a RAW writer (bypasses SetState entirely; invisible to the
indicator-level trap) resets the cell between damage events -- under the
bench's 1 Hz metronome harness it restarted the sequence before the 1.8 s
voice note; sporadic real-play damage is unaffected (one edge -> full
sequence). Needs a cdb write-watchpoint session; candidates: a recon raw
+0x2c-equivalent write or a struct copy spanning it.
Also decoded en route: the AnimationState triggers on the limp states play
EngineShiftRev01 (the downshift foley) -- working, and NOT the voice.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Correcting my own framing in 7e57816. The dropout players actually reported was
fixed on 2026-07-23 by a999e5c, and nobody has complained since. I read the
leftover ACQUIRE FAILED lines in the night-5 logs as the same live bug and
raised the ceiling by default. They are not the same bug.
What is still in the logs is transient exhaustion, not a leak: the census sits
at 45-60 live, spikes during a firefight, and drains straight back (226 -> 42),
which is the priority steal loop working. A voice dropped in that window is
competing with ~256 already sounding, so it is very likely sub-perceptual --
which is consistent with the thing nobody is reporting.
And raising the cap is not free. The 256 ceiling doubles as a governor: the
steal loop only steals when the incoming source outranks a running one, so a
higher cap means many more voices mixing simultaneously, and with EFX reverb and
the lowpass chains live that is real CPU -- spent precisely during heavy combat,
when the frame budget is already tightest. I verified the ceiling moves. I never
measured frame time under a real firefight, and a solo bench cannot produce one.
Shipping that to testers tonight would be gambling their framerate against a
complaint nobody is making.
So the knob stays and the default does not move. BT_AUDIO_SOURCES=<n> raises it
with no rebuild if dropouts are ever reported again -- measure frame time while
you do. Unset reports `granted mono=255 stereo=1`, exactly what testers have
been running.
Keeping the diagnostic read-back either way, since it earned its keep: asking
for 64 grants 240, because OpenAL Soft has a floor of its own, which is also why
the NULL default lands on 256. A request is not a promise.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Field report from 2026-07-23 was "audio cutting in and out toward the end of the
match". The night-5 logs say it precisely: 2017 ACQUIRE FAILED lines in a single
match, and every last one of them at live=256.
It is not a leak, which is where I started and where the code comments still
pointed. The source census sits at a healthy 45-60 live, spikes during
firefights, and drains right back down afterwards -- 226 back to 42 in one
window -- so the engine's priority steal loop is doing exactly what it should:
AudioSourceStop/SuspendMaintenance -> ReleaseChannels -> ReleaseSourceSet ->
alDeleteSources, with the live counter following it down. The 2026-07-23 fix
(a999e5c, deleting sources one at a time instead of the spec-atomic bulk call)
was real and is what makes that drain work. It just was not the ceiling.
The ceiling was ours. MakeAudioRenderer called alcCreateContext(device, NULL) --
no attribute list at all -- so OpenAL Soft applied its default budget of 256
mono sources. That number has nothing to do with the 1995 audio hardware; it is
just what you get for not asking. Meanwhile a busy match logs about 7200
explosions, each spawning three DPLIndependantEffect voices, so the pool pins at
the cap and every acquire during that window fails outright and drops its sound.
The peak we actually observed was 226 against a 256 cap, i.e. the bursts were
already scraping the ceiling.
So ask. ALC_MONO_SOURCES at 1024 by default, BT_AUDIO_SOURCES to override for
A/B testing, and read back what the driver GRANTED rather than assuming the
request was honoured. That last part earns its keep: asking for 64 grants 240,
because OpenAL Soft has a floor of its own -- which is also the reason the NULL
default landed on 256 in the first place.
requested 1024 -> granted 1024
requested 2048 -> granted 2048
requested 64 -> granted 240 (driver floor)
Costs mixing headroom, not hardware voices -- OpenAL Soft mixes in software and
only touches voices that are actually sounding.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Field report: audio cutting in and out toward the end of the match.
Three structural defects in the port's source-pool lifecycle:
1. SourceSet.sources[] was uninitialized heap garbage until first
acquisition -- AL names are small ints, so a recycled-heap slot could
alias a LIVE source owned by another sound; alIsSource() skipped
generation, two sounds shared one source, and whichever released
first deleted the other's mid-play. SourceSet ctor now zero-inits.
2. ReleaseSourceSet's bulk alDeleteSources(count, sources) is ATOMIC on
invalid names (AL spec) -- one empty/-1 slot and NOTHING deleted;
after the first pool exhaustion every partial release leaked and the
pool never recovered. Now per-slot guarded delete, slots parked at 0.
3. Failed acquisitions stranded partial sets forever (dropped transients
are never released by anything). ReleaseChannels() handback at both
failure sites (StartRequest + dormant-resume) + a dtor backstop.
Verified: 2-node MP smoke (mission + lobby-loop relaunch clean, census
live/acquireFails 0/0) + 100s solo combat smoke (52 audio triggers,
zero AL errors). Awaiting live playtest confirmation at scale.
KB: context/wintesla-port.md audio-dropout section.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Field report (playtest night + operator): audio cutting in and out
toward the end of the match -- the classic shape of source-pool
pressure (sources release only at entity teardown; long matches
accumulate looping occupants like wreck burn/smoke until transients
lose the voice fight; OpenAL mixing capacity is finite). L4AUDRND now
tracks live/deleted/failed sources: a 30s '[audio] source census' line
plus an ALWAYS-logged line on every acquisition failure (the smoking
gun). Soak test next to confirm the leak/occupation slope.
Also: the ffa3933 jam-entry log had a raw newline in a string literal
(committed unbuilt -- process violation, caught by this build).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The game audio plays hot with no in-game control -- testers couldn't
hear voice chat without the Windows mixer (playtest night). The -/=
keys now step the OpenAL listener gain (the master scale every source
inherits) in 5% steps, clamped 0..150%, edge-detected in the per-frame
poll with a foreground guard; the value persists to content\volume.cfg
and reloads at boot (BT_AUDIO_VOLUME env still wins when set; default
stays 0.6). README updated.
Verified: volume.cfg 0.25 -> boot log "master gain=0.25"; key stepping
needs a live focused session (awaiting live verification).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The weekend's crash family root-caused under cdb: SourceSet.sources[5]
receives the AllExplosion death preset's 25 streamed voices --
RequestAudioChannels wrote 20 OpenAL source ids past the array, smashing
the neighbouring heap object (Release: a vtable overwritten with a source
id -> delayed silent AV at AudioControlEvent::Send; Debug: CRT heap-
corruption abort in the death Explosion's audio teardown). Explains all
three #12 crash flavors (solo enemy kill -- stack-confirmed; MP self-
death; MP peer PEER_DOWN cascade -- same generic teardown).
Fix: sources[] sized to AUDIO_SOURCESET_CAPACITY=25, static_assert
lockstep with MAX_PRESET_SAMPLES, + defence-in-depth clamps at the ctor
and acquisition sites. Soak: 26+ deaths across glass AND pod builds
under cdb, zero faults, full wreck lifecycle every time. Gotchas S21:
the fixed-array-vs-streamed-count overflow class + sweep note.
Awaiting human verification: the MP death-and-survive session.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New L4AUDEFX bridge (OpenAL Soft ALC_EXT_EFX): one EAXReverb aux slot at the
authentic AUDIO.INI global_reverb_scale (0.3) + a scratch AL_FILTER_LOWPASS
(params copied at attach).
F9 filters (was: computed then thrown away -- everything full-bright at all
distances): Dynamic3D ExecuteModel drives GAINHF from highFreqCutoffScale x
brightnessScale x maxMIDIFilterCutoff, UNGATED (decomp part_008.c:7496,
7589-7604 -- every moving 3D sound dulls with distance per the AUDIO.INI
knee-60/exp-2.0 model); Static3D from brightness x max (:7831-7884); Direct
inside its existing gated NRPN-rate block. AWE 100-8000 Hz curve -> EFX
5 kHz-reference gainhf via a 2-pole approximation [T3 curve, endpoints exact].
F11 reverb (was: bone-dry everywhere): 3D patch sources attach an aux send at
Start, exactly where the original sent CC91 = global_reverb_scale
(part_008.c:7278-7394); Direct cockpit sources keep CC91=0 -- dry. The
wet-exterior vs dry-cockpit contrast is back.
F12 placement (was: every cockpit sound dead-center): DirectPatchSource
Start places sources by the authored 6-value position enum (front/rear card
+ pan CC10, decomp @00463848/@004638a8) as listener-relative directions,
composed with the zone L/C/R pan. New PatchResource GetBankID/GetPatchID
pass-throughs (the LOD accessor is protected).
Regression (35s drive+fire): EFX READY, stable, deliveries unregressed, no AL
errors.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The raw AWE32 samples are hot and everything plays at full per-source gain, so the
mix was loud. Set alListenerf(AL_GAIN) once at device init -- it scales EVERY
source (master volume). Default 0.6; override with BT_AUDIO_VOLUME=<0.0..1.0+>
(0 = mute, 1 = full, >1 = boost).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
User reported never hearing any audio. Root-caused: the KB "Audio: DONE" was wrong.
The OpenAL playback chain IS fully implemented (device/buffers/sources; PlayNote
really calls alSourcePlay), but three gaps kept it silent:
1. GATED OFF (the "no sound" root cause): BTL4Application::MakeAudioRenderer returned
NULL unless the pod's AWE_FRONT/AWE_REAR AWE32-card env vars were set -- authentic
1995 pod behavior, dead on modern hardware, so the renderer was NEVER created.
FIXED: default audio ON (BT_NO_AUDIO=1 restores silence; AWE vars still force-on).
Verified: the OpenAL device now opens with no env vars ([audio] device OPENED).
2. Soundbank STUB: allPresets[2][100] (btstubs.cpp) is zero-init -> PRESET_isImplemented
false -> 0 buffers load. Sample data exists (AUDIO1/2.RES: EnginePower/LaserAFire/...)
but the event->sample map is gone (not in the decomp). STILL OPEN.
3. No triggering: the reconstructed game never creates AudioEntities on events. STILL OPEN.
So the device opens but nothing loads/plays yet -- enabling real sound needs the
soundbank reconstructed + the game triggers wired (a proper audio wave). Added a
BT_AUDIO_LOG trace harness across L4AUDRND/L4AUDRES/L4AUDLVL. KB corrected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Clean, self-contained extraction of the BattleTech-specific work from the
reverse-engineering workspace -- engine + game + content + build, with nothing
from Red Planet or the raw archive dumps. Builds green (Win32) and runs the
single-player drive->animate->target->fire->damage->destroy loop out of the box.
Layout:
engine/ MUNGA + MUNGA_L4 shared 2007 engine, carrying our BT render/loader
work (bgfload/L4D3D/L4VIDEO: BSL bit-slice decode, LOD/ground/shadow
models) + image codec; the minimal rp/ headers the audio HAL needs
game/ reconstructed BT logic + surviving-original BT source + fwd shims
+ WinMain launcher
content/ full runtime tree (BTL4.RES, VIDEO/, GAUGE/, AUDIO/, eggs, BTDPL.INI)
docs/ format specs + reconstruction ledgers
reference/ raw Ghidra pseudocode (recon source-of-truth) + decomp exporter
tools/ MP console emulator + map/resource scanners
One top-level CMake builds munga_engine lib + bt410_l4 game lib + btl4.exe.
All paths relativized (186 fwd shims + ~437 CMake abs paths -> repo-relative);
DXSDK is the one external, overridable via -DDXSDK. Verified: builds to a
byte-identical 2.27MB exe and runs combat (TARGET DESTROYED, 0 crashes) against
the bundled content.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
