From d9b5755713fe2f208793991c1c47f0e6498bac78 Mon Sep 17 00:00:00 2001
From: Florian Will <florian.will@gmail.com>
Date: Fri, 30 Jun 2023 09:25:14 +0200
Subject: [PATCH] mp4a: Support QT sound sample description version 1.

If the QT version in a sound sample description is set to 1, there are 4
additional 32-bit integers in the description record according to QT
docs.

At least some recently sold iPhones produce sound sample description
records like this, so this change fixes a parsing failure for these
files.
---
 src/mp4box/mp4a.rs | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/src/mp4box/mp4a.rs b/src/mp4box/mp4a.rs
index e4d6a45..0dcfca7 100644
--- a/src/mp4box/mp4a.rs
+++ b/src/mp4box/mp4a.rs
@@ -83,12 +83,22 @@ impl<R: Read + Seek> ReadBox<&mut R> for Mp4aBox {
         reader.read_u16::<BigEndian>()?; // reserved
         let data_reference_index = reader.read_u16::<BigEndian>()?;
 
-        reader.read_u64::<BigEndian>()?; // reserved
+        // Next 8 bytes should be all 0 in ISO MP4
+        let qt_version = reader.read_u16::<BigEndian>()?; // QT: 0, 1 or 2
+        reader.read_u16::<BigEndian>()?; // QT: revision level, should be 0
+        reader.read_u32::<BigEndian>()?; // QT: vendor, should be 0
+
         let channelcount = reader.read_u16::<BigEndian>()?;
         let samplesize = reader.read_u16::<BigEndian>()?;
         reader.read_u32::<BigEndian>()?; // pre-defined, reserved
         let samplerate = FixedPointU16::new_raw(reader.read_u32::<BigEndian>()?);
 
+        if qt_version == 1 {
+            // Samples per packet, bytes per packet,
+            // bytes per frame, bytes per sample
+            skip_bytes(reader, 4 * 4)?;
+        }
+
         let mut esds = None;
         let current = reader.stream_position()?;
         if current < start + size {