PR #21075 opened by tangsha
URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21075
Patch URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21075.patch

The this patch is based on the latest master branch (not a release branch) – 
I’ve synced with the mainline before submitting the changes.
Attached is error.wav – this is the problematic audio file that gets fixed by 
the patch. Prior to the change, decoding this file resulted in [briefly 
describe the issue, e.g., "distorted audio, incorrect step index calculations, 
or data misalignment"]; after applying the patch, the audio decodes correctly.
 
Regarding the skipped byte: Based on analyzing the ADPCM format of this sample, 
the skipped byte is padding/reserved space (not part of the step_index field). 
The original code incorrectly read this padding byte as part of a 16-bit 
step_index, leading to the decoding error in error.wav.


From ec27e082181600f31c1aaad3e565853cfa81daaa Mon Sep 17 00:00:00 2001
From: tangsha <[email protected]>
Date: Mon, 1 Dec 2025 16:15:35 +0800
Subject: [PATCH] avcodec/adpcm: Fix the decoding error caused by incorrect

---
 libavcodec/adpcm.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/libavcodec/adpcm.c b/libavcodec/adpcm.c
index 72fbb841a4..7aac254a4c 100644
--- a/libavcodec/adpcm.c
+++ b/libavcodec/adpcm.c
@@ -1512,7 +1512,8 @@ static int adpcm_decode_frame(AVCodecContext *avctx, 
AVFrame *frame,
             ADPCMChannelStatus *cs = &c->status[i];
             cs->predictor = samples_p[i][0] = 
sign_extend(bytestream2_get_le16u(&gb), 16);
 
-            cs->step_index = sign_extend(bytestream2_get_le16u(&gb), 16);
+           cs->step_index = bytestream2_get_byteu(&gb);
+           bytestream2_skip(&gb, sizeof(uint8_t));
             if (cs->step_index > 88u){
                 av_log(avctx, AV_LOG_ERROR, "ERROR: step_index[%d] = %i\n",
                        i, cs->step_index);
-- 
2.49.1

_______________________________________________
ffmpeg-devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to