Steve Lhomme pushed to branch master at VideoLAN / VLC
Commits:
2f1601e0 by Martin Finkel at 2026-01-09T11:43:23+00:00
Revert "demux: libmp4: read Qt double sample rate in one call"
This reverts commit 6aab1e0b2b25971eb2e217c93b68a5ca831f5334.
MP4_COPY_BYTES does a raw memcpy without endianness conversion.
QuickTime Sound Description V2 stores the sample rate as a big-endian
64-bit IEEE 754 double. On little-endian systems (Windows, Android ARM,
etc.), this results in a garbage sample rate value, causing PCM audio
playback to fail with "bad samplerate: 0 Hz".
The original code using MP4_GET8BYTES + memcpy correctly handles the
big-endian to native conversion.
- - - - -
1 changed file:
- modules/demux/mp4/libmp4.c
Changes:
=====================================
modules/demux/mp4/libmp4.c
=====================================
@@ -2698,6 +2698,7 @@ static int MP4_ReadBox_sample_soun( stream_t *p_stream,
MP4_Box_t *p_box )
{
/* SoundDescriptionV2 */
double f_sample_rate;
+ int64_t i_dummy64;
uint32_t i_channel, i_extoffset, i_dummy32;
/* Checks */
@@ -2714,7 +2715,8 @@ static int MP4_ReadBox_sample_soun( stream_t *p_stream,
MP4_Box_t *p_box )
/* !Checks */
MP4_GET4BYTES( i_extoffset ); /* offset to stsd extensions */
- MP4_COPY_BYTES( f_sample_rate, 8 );
+ MP4_GET8BYTES( i_dummy64 );
+ memcpy( &f_sample_rate, &i_dummy64, 8 );
msg_Dbg( p_stream, "read box: %f Hz", f_sample_rate );
/* Rounding error with lo, but we don't care since we do not support
fractional audio rate */
p_box->data.p_sample_soun->i_sampleratehi = (uint16_t)f_sample_rate;
View it on GitLab:
https://code.videolan.org/videolan/vlc/-/commit/2f1601e03a40137f9a2168249c420a508733a444
--
View it on GitLab:
https://code.videolan.org/videolan/vlc/-/commit/2f1601e03a40137f9a2168249c420a508733a444
You're receiving this email because of your account on code.videolan.org.
VideoLAN code repository instance_______________________________________________
vlc-commits mailing list
[email protected]
https://mailman.videolan.org/listinfo/vlc-commits
