From 8c922abd7747387bf3a864b9781cc164acd08934 Mon Sep 17 00:00:00 2001
From: Dipshet <264011288+Dipshet@users.noreply.github.com>
Date: Fri, 31 Jul 2026 04:23:20 +0200
Subject: [PATCH] Preserve the XMA stream timeline across undecodable frames

A frame the decoder backend rejects advances the read offset but emitted nothing, silently shortening that stream by 512 samples, permanent desync for multi-stream sources (5.1 premixes as parallel stereo XMA). Emit the frame as silence instead so the timeline holds. Robustness: no in-game trigger known in AC6. audio_xma_preserve_timeline (default on).
---
 .../src/native/audio/audio_system.cpp         | 10 +++++++
 .../src/native/audio/xma/context.cpp          | 28 +++++++++++++++++--
 2 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/thirdparty/rexglue-sdk/src/native/audio/audio_system.cpp b/thirdparty/rexglue-sdk/src/native/audio/audio_system.cpp
index bb1b29e6..10c5f301 100644
--- a/thirdparty/rexglue-sdk/src/native/audio/audio_system.cpp
+++ b/thirdparty/rexglue-sdk/src/native/audio/audio_system.cpp
@@ -19,6 +19,16 @@ REXCVAR_DEFINE_BOOL(audio_trace_render_driver_verbose, false, "Audio",
                     "Trace render-driver activity");
 REXCVAR_DEFINE_BOOL(audio_deep_trace, false, "Audio",
                     "Enable verbose runtime audio tracing");
+REXCVAR_DEFINE_BOOL(audio_xma_preserve_timeline, true, "Audio",
+                    "When the XMA decoder backend rejects a damaged frame, emit that "
+                    "frame's worth of silence instead of dropping it. The input read "
+                    "offset advances past the frame either way, so dropping the output "
+                    "silently shortens the stream's timeline by 512 samples per damaged "
+                    "frame - which would permanently desynchronize multi-stream sources "
+                    "(5.1 premixes carried as parallel stereo streams, e.g. AC6's "
+                    "cutscene mixes). Robustness fix - no in-game trigger is known in "
+                    "AC6 (instrumented runs decode every frame). On by default; off "
+                    "restores the legacy drop behavior.");
 
 namespace rex::audio {
 
diff --git a/thirdparty/rexglue-sdk/src/native/audio/xma/context.cpp b/thirdparty/rexglue-sdk/src/native/audio/xma/context.cpp
index 73994824..6550bc44 100644
--- a/thirdparty/rexglue-sdk/src/native/audio/xma/context.cpp
+++ b/thirdparty/rexglue-sdk/src/native/audio/xma/context.cpp
@@ -17,6 +17,7 @@
 #include <native/stream.h>
 
 REXCVAR_DECLARE(bool, audio_deep_trace);
+REXCVAR_DECLARE(bool, audio_xma_preserve_timeline);
 
 namespace rex::audio {
 
@@ -853,11 +854,32 @@ void XmaContext::Decode(XMA_CONTEXT_DATA* data) {
       .sample_rate = static_cast<uint32_t>(GetSampleRate(data->sample_rate)),
       .is_two_channel = bool(data->is_stereo),
   };
-  if (decoder_backend_ &&
+  const bool frame_decoded =
+      decoder_backend_ &&
       decoder_backend_->DecodePacket(
-          decode_request, std::span<uint8_t>(raw_frame_.data(), raw_frame_.size()))) {
+          decode_request, std::span<uint8_t>(raw_frame_.data(), raw_frame_.size()));
+  if (!frame_decoded && REXCVAR_GET(audio_xma_preserve_timeline)) {
+    // A frame the decoder rejects still occupies exactly kSamplesPerFrame
+    // samples of the stream's timeline, and the read offset advances past it
+    // below either way. Dropping the output entirely (the legacy behavior)
+    // silently shortens the stream by one frame - a latent correctness bug
+    // for multi-stream sources that must stay sample-locked, e.g. 5.1
+    // premixes carried as parallel stereo streams (AC6's cutscene mixes).
+    // Instrumented AC6 runs decode every cutscene frame successfully, so no
+    // in-game trigger is known - this is robustness against decode errors,
+    // not a fix for an audible symptom. Deliver the frame as silence
+    // instead - raw_frame_ is zero-filled above, so falling through emits
+    // one frame of silence in this stream (~10 ms, masked by the other
+    // streams) and the timeline holds.
+    REXAPU_DEBUG(
+        "XmaContext {}: undecodable frame (packet={} offset={} frame_bits={}) - "
+        "emitting silence to preserve the stream timeline",
+        id(), last_packet_index_, last_input_read_offset_before_,
+        packet_info.current_frame_size_);
+  }
+  if (frame_decoded || REXCVAR_GET(audio_xma_preserve_timeline)) {
     current_frame_remaining_subframes_ = 4 << data->is_stereo;
-    last_decode_succeeded_ = true;
+    last_decode_succeeded_ = frame_decoded;
 
     if (is_loop_end_frame) {
       loop_frame_output_limit_ = (data->loop_subframe_end + 1) << data->is_stereo;