mp4/fmp4: support flac inside the iso (f)mp4 container

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/1401>

docs/plugins/gst_plugins_cache.json

@@ -2038,7 +2038,7 @@
                 "long-name": "ISOFMP4Mux",
                 "pad-templates": {
                     "sink_%%u": {
-                        "caps": "video/x-h264:\n  stream-format: { (string)avc, (string)avc3 }\n      alignment: au\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-h265:\n  stream-format: { (string)hvc1, (string)hev1 }\n      alignment: au\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-vp8:\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-vp9:\n        profile: { (string)0, (string)1, (string)2, (string)3 }\n  chroma-format: { (string)4:2:0, (string)4:2:2, (string)4:4:4 }\n bit-depth-luma: { (uint)8, (uint)10, (uint)12 }\nbit-depth-chroma: { (uint)8, (uint)10, (uint)12 }\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-av1:\n  stream-format: obu-stream\n      alignment: tu\n        profile: { (string)main, (string)high, (string)professional }\n  chroma-format: { (string)4:0:0, (string)4:2:0, (string)4:2:2, (string)4:4:4 }\n bit-depth-luma: { (uint)8, (uint)10, (uint)12 }\nbit-depth-chroma: { (uint)8, (uint)10, (uint)12 }\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\naudio/mpeg:\n    mpegversion: 4\n  stream-format: raw\n       channels: [ 1, 65535 ]\n           rate: [ 1, 2147483647 ]\naudio/x-opus:\nchannel-mapping-family: [ 0, 255 ]\n       channels: [ 1, 8 ]\n           rate: [ 1, 2147483647 ]\n",
+                        "caps": "video/x-h264:\n  stream-format: { (string)avc, (string)avc3 }\n      alignment: au\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-h265:\n  stream-format: { (string)hvc1, (string)hev1 }\n      alignment: au\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-vp8:\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-vp9:\n        profile: { (string)0, (string)1, (string)2, (string)3 }\n  chroma-format: { (string)4:2:0, (string)4:2:2, (string)4:4:4 }\n bit-depth-luma: { (uint)8, (uint)10, (uint)12 }\nbit-depth-chroma: { (uint)8, (uint)10, (uint)12 }\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-av1:\n  stream-format: obu-stream\n      alignment: tu\n        profile: { (string)main, (string)high, (string)professional }\n  chroma-format: { (string)4:0:0, (string)4:2:0, (string)4:2:2, (string)4:4:4 }\n bit-depth-luma: { (uint)8, (uint)10, (uint)12 }\nbit-depth-chroma: { (uint)8, (uint)10, (uint)12 }\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\naudio/mpeg:\n    mpegversion: 4\n  stream-format: raw\n       channels: [ 1, 65535 ]\n           rate: [ 1, 2147483647 ]\naudio/x-opus:\nchannel-mapping-family: [ 0, 255 ]\n       channels: [ 1, 8 ]\n           rate: [ 1, 2147483647 ]\naudio/x-flac:\n         framed: true\n       channels: [ 1, 8 ]\n           rate: [ 1, 655350 ]\n",
                         "direction": "sink",
                         "presence": "request",
                         "type": "GstFMP4MuxPad"
@@ -3318,7 +3318,7 @@
                 "klass": "Codec/Muxer",
                 "pad-templates": {
                     "sink_%%u": {
-                        "caps": "video/x-h264:\n  stream-format: { (string)avc, (string)avc3 }\n      alignment: au\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-h265:\n  stream-format: { (string)hvc1, (string)hev1 }\n      alignment: au\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-vp8:\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-vp9:\n        profile: { (string)0, (string)1, (string)2, (string)3 }\n  chroma-format: { (string)4:2:0, (string)4:2:2, (string)4:4:4 }\n bit-depth-luma: { (uint)8, (uint)10, (uint)12 }\nbit-depth-chroma: { (uint)8, (uint)10, (uint)12 }\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-av1:\n  stream-format: obu-stream\n      alignment: tu\n        profile: { (string)main, (string)high, (string)professional }\n  chroma-format: { (string)4:0:0, (string)4:2:0, (string)4:2:2, (string)4:4:4 }\n bit-depth-luma: { (uint)8, (uint)10, (uint)12 }\nbit-depth-chroma: { (uint)8, (uint)10, (uint)12 }\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\naudio/mpeg:\n    mpegversion: 4\n  stream-format: raw\n       channels: [ 1, 65535 ]\n           rate: [ 1, 2147483647 ]\naudio/x-opus:\nchannel-mapping-family: [ 0, 255 ]\n       channels: [ 1, 8 ]\n           rate: [ 1, 2147483647 ]\n",
+                        "caps": "video/x-h264:\n  stream-format: { (string)avc, (string)avc3 }\n      alignment: au\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-h265:\n  stream-format: { (string)hvc1, (string)hev1 }\n      alignment: au\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-vp8:\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-vp9:\n        profile: { (string)0, (string)1, (string)2, (string)3 }\n  chroma-format: { (string)4:2:0, (string)4:2:2, (string)4:4:4 }\n bit-depth-luma: { (uint)8, (uint)10, (uint)12 }\nbit-depth-chroma: { (uint)8, (uint)10, (uint)12 }\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\nvideo/x-av1:\n  stream-format: obu-stream\n      alignment: tu\n        profile: { (string)main, (string)high, (string)professional }\n  chroma-format: { (string)4:0:0, (string)4:2:0, (string)4:2:2, (string)4:4:4 }\n bit-depth-luma: { (uint)8, (uint)10, (uint)12 }\nbit-depth-chroma: { (uint)8, (uint)10, (uint)12 }\n          width: [ 1, 65535 ]\n         height: [ 1, 65535 ]\naudio/mpeg:\n    mpegversion: 4\n  stream-format: raw\n       channels: [ 1, 65535 ]\n           rate: [ 1, 2147483647 ]\naudio/x-opus:\nchannel-mapping-family: [ 0, 255 ]\n       channels: [ 1, 8 ]\n           rate: [ 1, 2147483647 ]\naudio/x-flac:\n         framed: true\n       channels: [ 1, 8 ]\n           rate: [ 1, 655350 ]\n",
                         "direction": "sink",
                         "presence": "request",
                         "type": "GstRsMP4MuxPad"

mux/fmp4/src/fmp4mux/boxes.rs

@@ -9,6 +9,7 @@
 use gst::prelude::*;
 
 use anyhow::{anyhow, bail, Context, Error};
+use std::convert::TryFrom;
 
 use super::Buffer;
 
@@ -604,9 +605,8 @@ fn write_tkhd(
     // Volume
     let s = stream.caps.structure(0).unwrap();
     match s.name().as_str() {
-        "audio/mpeg" | "audio/x-opus" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
+        "audio/mpeg" | "audio/x-opus" | "audio/x-flac" | "audio/x-alaw" | "audio/x-mulaw"
-            v.extend((1u16 << 8).to_be_bytes())
+        | "audio/x-adpcm" => v.extend((1u16 << 8).to_be_bytes()),
-        }
         _ => v.extend(0u16.to_be_bytes()),
     }
 
@@ -745,9 +745,8 @@ fn write_hdlr(
     let (handler_type, name) = match s.name().as_str() {
         "video/x-h264" | "video/x-h265" | "video/x-vp8" | "video/x-vp9" | "video/x-av1"
         | "image/jpeg" => (b"vide", b"VideoHandler\0".as_slice()),
-        "audio/mpeg" | "audio/x-opus" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
+        "audio/mpeg" | "audio/x-opus" | "audio/x-flac" | "audio/x-alaw" | "audio/x-mulaw"
-            (b"soun", b"SoundHandler\0".as_slice())
+        | "audio/x-adpcm" => (b"soun", b"SoundHandler\0".as_slice()),
-        }
         "application/x-onvif-metadata" => (b"meta", b"MetadataHandler\0".as_slice()),
         _ => unreachable!(),
     };
@@ -777,7 +776,8 @@ fn write_minf(
             // Flags are always 1 for unspecified reasons
             write_full_box(v, b"vmhd", FULL_BOX_VERSION_0, 1, |v| write_vmhd(v, cfg))?
         }
-        "audio/mpeg" | "audio/x-opus" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
+        "audio/mpeg" | "audio/x-opus" | "audio/x-flac" | "audio/x-alaw" | "audio/x-mulaw"
+        | "audio/x-adpcm" => {
             write_full_box(v, b"smhd", FULL_BOX_VERSION_0, FULL_BOX_FLAGS_NONE, |v| {
                 write_smhd(v, cfg)
             })?
@@ -886,9 +886,8 @@ fn write_stsd(
     match s.name().as_str() {
         "video/x-h264" | "video/x-h265" | "video/x-vp8" | "video/x-vp9" | "video/x-av1"
         | "image/jpeg" => write_visual_sample_entry(v, cfg, stream)?,
-        "audio/mpeg" | "audio/x-opus" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
+        "audio/mpeg" | "audio/x-opus" | "audio/x-flac" | "audio/x-alaw" | "audio/x-mulaw"
-            write_audio_sample_entry(v, cfg, stream)?
+        | "audio/x-adpcm" => write_audio_sample_entry(v, cfg, stream)?,
-        }
         "application/x-onvif-metadata" => write_xml_meta_data_sample_entry(v, cfg, stream)?,
         _ => unreachable!(),
     }
@@ -1262,6 +1261,7 @@ fn write_audio_sample_entry(
     let fourcc = match s.name().as_str() {
         "audio/mpeg" => b"mp4a",
         "audio/x-opus" => b"Opus",
+        "audio/x-flac" => b"fLaC",
         "audio/x-alaw" => b"alaw",
         "audio/x-mulaw" => b"ulaw",
         "audio/x-adpcm" => {
@@ -1280,6 +1280,10 @@ fn write_audio_sample_entry(
             let bitrate = s.get::<i32>("bitrate").context("no ADPCM bitrate field")?;
             (bitrate / 8000) as u16
         }
+        "audio/x-flac" => with_flac_metadata(&stream.caps, |streaminfo, _| {
+            1 + (u16::from_be_bytes([streaminfo[16], streaminfo[17]]) >> 4 & 0b11111)
+        })
+        .context("FLAC metadata error")?,
         _ => 16u16,
     };
 
@@ -1322,6 +1326,9 @@ fn write_audio_sample_entry(
             "audio/x-opus" => {
                 write_dops(v, &stream.caps)?;
             }
+            "audio/x-flac" => {
+                write_dfla(v, &stream.caps)?;
+            }
             "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
                 // Nothing to do here
             }
@@ -1516,6 +1523,35 @@ fn write_dops(v: &mut Vec<u8>, caps: &gst::Caps) -> Result<(), Error> {
     })
 }
 
+fn with_flac_metadata<R>(
+    caps: &gst::Caps,
+    cb: impl FnOnce(&[u8], &[gst::glib::SendValue]) -> R,
+) -> Result<R, Error> {
+    let caps = caps.structure(0).unwrap();
+    let header = caps.get::<gst::ArrayRef>("streamheader").unwrap();
+    let (streaminfo, remainder) = header.as_ref().split_first().unwrap();
+    let streaminfo = streaminfo.get::<&gst::BufferRef>().unwrap();
+    let streaminfo = streaminfo.map_readable().unwrap();
+    // 13 bytes for the Ogg/FLAC prefix and 38 for the streaminfo itself.
+    match <&[_; 13 + 38]>::try_from(streaminfo.as_slice()) {
+        Ok(i) if i.starts_with(b"\x7FFLAC\x01\x00") => Ok(cb(&i[13..], remainder)),
+        Ok(_) | Err(_) => bail!("Unknown streamheader format"),
+    }
+}
+
+fn write_dfla(v: &mut Vec<u8>, caps: &gst::Caps) -> Result<(), Error> {
+    write_full_box(v, b"dfLa", 0, 0, move |v| {
+        with_flac_metadata(caps, |streaminfo, remainder| {
+            v.extend(streaminfo);
+            for metadata in remainder {
+                let metadata = metadata.get::<&gst::BufferRef>().unwrap();
+                let metadata = metadata.map_readable().unwrap();
+                v.extend(&metadata[..]);
+            }
+        })
+    })
+}
+
 fn write_xml_meta_data_sample_entry(
     v: &mut Vec<u8>,
     _cfg: &super::HeaderConfiguration,

mux/fmp4/src/fmp4mux/imp.rs

@@ -205,6 +205,8 @@ struct Stream {
     caps: gst::Caps,
     /// Whether this stream is intra-only and has frame reordering.
     delta_frames: DeltaFrames,
+    /// Whether this stream might have header frames without timestamps that should be ignored.
+    discard_header_buffers: bool,
 
     /// Currently queued GOPs, including incomplete ones.
     queued_gops: VecDeque<Gop>,
@@ -271,11 +273,17 @@ pub(crate) struct FMP4Mux {
 
 impl FMP4Mux {
     /// Checks if a buffer is valid according to the stream configuration.
-    fn check_buffer(
+    fn check_buffer(buffer: &gst::BufferRef, stream: &Stream) -> Result<(), gst::FlowError> {
-        buffer: &gst::BufferRef,
+        let Stream {
-        sinkpad: &super::FMP4MuxPad,
+            sinkpad,
-        delta_frames: super::DeltaFrames,
+            delta_frames,
-    ) -> Result<(), gst::FlowError> {
+            discard_header_buffers,
+            ..
+        } = stream;
+        if *discard_header_buffers && buffer.flags().contains(gst::BufferFlags::HEADER) {
+            return Err(gst_base::AGGREGATOR_FLOW_NEED_DATA);
+        }
+
         if delta_frames.requires_dts() && buffer.dts().is_none() {
             gst::error!(CAT, obj: sinkpad, "Require DTS for video streams");
             return Err(gst::FlowError::Error);
@@ -314,12 +322,10 @@ impl FMP4Mux {
         }
 
         // Pop buffer here, it will be stored in the pre-queue after calculating its timestamps
-        let mut buffer = match stream.sinkpad.pop_buffer() {
+        let Some(mut buffer) = stream.sinkpad.pop_buffer() else {
-            None => return Ok(None),
+            return Ok(None);
-            Some(buffer) => buffer,
         };
-
+        Self::check_buffer(&buffer, stream)?;
-        Self::check_buffer(&buffer, &stream.sinkpad, stream.delta_frames)?;
 
         let segment = match stream.sinkpad.segment().downcast::<gst::ClockTime>().ok() {
             Some(segment) => segment,
@@ -2555,6 +2561,7 @@ impl FMP4Mux {
             let s = caps.structure(0).unwrap();
 
             let mut delta_frames = DeltaFrames::IntraOnly;
+            let mut discard_header_buffers = false;
             match s.name().as_str() {
                 "video/x-h264" | "video/x-h265" => {
                     if !s.has_field_with_type("codec_data", gst::Buffer::static_type()) {
@@ -2598,6 +2605,13 @@ impl FMP4Mux {
                         return Err(gst::FlowError::NotNegotiated);
                     }
                 }
+                "audio/x-flac" => {
+                    discard_header_buffers = true;
+                    if let Err(e) = s.get::<gst::ArrayRef>("streamheader") {
+                        gst::error!(CAT, obj: pad, "Muxing FLAC into MP4 needs streamheader: {}", e);
+                        return Err(gst::FlowError::NotNegotiated);
+                    };
+                }
                 "audio/x-alaw" | "audio/x-mulaw" => (),
                 "audio/x-adpcm" => (),
                 "application/x-onvif-metadata" => (),
@@ -2608,6 +2622,7 @@ impl FMP4Mux {
                 sinkpad: pad,
                 caps,
                 delta_frames,
+                discard_header_buffers,
                 pre_queue: VecDeque::new(),
                 queued_gops: VecDeque::new(),
                 fragment_filled: false,
@@ -3465,6 +3480,11 @@ impl ElementImpl for ISOFMP4Mux {
                         .field("channels", gst::IntRange::new(1i32, 8))
                         .field("rate", gst::IntRange::new(1, i32::MAX))
                         .build(),
+                    gst::Structure::builder("audio/x-flac")
+                        .field("framed", true)
+                        .field("channels", gst::IntRange::<i32>::new(1, 8))
+                        .field("rate", gst::IntRange::<i32>::new(1, 10 * u16::MAX as i32))
+                        .build(),
                 ]
                 .into_iter()
                 .collect::<gst::Caps>(),

mux/fmp4/tests/tests.rs

@@ -19,6 +19,33 @@ fn init() {
     });
 }
 
+fn to_completion(pipeline: &gst::Pipeline) {
+    pipeline
+        .set_state(gst::State::Playing)
+        .expect("Unable to set the pipeline to the `Playing` state");
+
+    for msg in pipeline.bus().unwrap().iter_timed(gst::ClockTime::NONE) {
+        use gst::MessageView;
+
+        match msg.view() {
+            MessageView::Eos(..) => break,
+            MessageView::Error(err) => {
+                panic!(
+                    "Error from {:?}: {} ({:?})",
+                    err.src().map(|s| s.path_string()),
+                    err.error(),
+                    err.debug()
+                );
+            }
+            _ => (),
+        }
+    }
+
+    pipeline
+        .set_state(gst::State::Null)
+        .expect("Unable to set the pipeline to the `Null` state");
+}
+
 fn test_buffer_flags_single_stream(cmaf: bool, set_dts: bool, caps: gst::Caps) {
     let mut h = if cmaf {
         gst_check::Harness::new("cmafmux")
@@ -1993,3 +2020,21 @@ fn test_chunking_single_stream_gops_after_fragment_end_after_next_chunk_end() {
     let ev = h.pull_event().unwrap();
     assert_eq!(ev.type_(), gst::EventType::Eos);
 }
+
+#[test]
+fn test_roundtrip_vp9_flac() {
+    init();
+
+    let pipeline = gst::parse::launch(
+        r#"
+        videotestsrc num-buffers=99 ! vp9enc ! vp9parse ! mux.
+        audiotestsrc num-buffers=149 ! flacenc ! flacparse ! mux.
+        isofmp4mux name=mux ! qtdemux name=demux
+        demux.audio_0 ! queue ! flacdec ! fakesink
+        demux.video_0 ! queue ! vp9dec ! fakesink
+        "#,
+    )
+    .unwrap();
+    let pipeline = pipeline.downcast().unwrap();
+    to_completion(&pipeline);
+}

mux/mp4/src/mp4mux/boxes.rs

@@ -9,7 +9,7 @@
 use gst::prelude::*;
 
 use anyhow::{anyhow, bail, Context, Error};
-
+use std::convert::TryFrom;
 use std::str::FromStr;
 
 fn write_box<T, F: FnOnce(&mut Vec<u8>) -> Result<T, Error>>(
@@ -382,9 +382,8 @@ fn write_tkhd(
     // Volume
     let s = stream.caps.structure(0).unwrap();
     match s.name().as_str() {
-        "audio/mpeg" | "audio/x-opus" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
+        "audio/mpeg" | "audio/x-opus" | "audio/x-flac" | "audio/x-alaw" | "audio/x-mulaw"
-            v.extend((1u16 << 8).to_be_bytes())
+        | "audio/x-adpcm" => v.extend((1u16 << 8).to_be_bytes()),
-        }
         _ => v.extend(0u16.to_be_bytes()),
     }
 
@@ -514,9 +513,8 @@ fn write_hdlr(
     let (handler_type, name) = match s.name().as_str() {
         "video/x-h264" | "video/x-h265" | "video/x-vp8" | "video/x-vp9" | "video/x-av1"
         | "image/jpeg" => (b"vide", b"VideoHandler\0".as_slice()),
-        "audio/mpeg" | "audio/x-opus" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
+        "audio/mpeg" | "audio/x-opus" | "audio/x-flac" | "audio/x-alaw" | "audio/x-mulaw"
-            (b"soun", b"SoundHandler\0".as_slice())
+        | "audio/x-adpcm" => (b"soun", b"SoundHandler\0".as_slice()),
-        }
         "application/x-onvif-metadata" => (b"meta", b"MetadataHandler\0".as_slice()),
         _ => unreachable!(),
     };
@@ -546,7 +544,8 @@ fn write_minf(
             // Flags are always 1 for unspecified reasons
             write_full_box(v, b"vmhd", FULL_BOX_VERSION_0, 1, |v| write_vmhd(v, header))?
         }
-        "audio/mpeg" | "audio/x-opus" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
+        "audio/mpeg" | "audio/x-opus" | "audio/x-flac" | "audio/x-alaw" | "audio/x-mulaw"
+        | "audio/x-adpcm" => {
             write_full_box(v, b"smhd", FULL_BOX_VERSION_0, FULL_BOX_FLAGS_NONE, |v| {
                 write_smhd(v, header)
             })?
@@ -703,9 +702,8 @@ fn write_stsd(
     match s.name().as_str() {
         "video/x-h264" | "video/x-h265" | "video/x-vp8" | "video/x-vp9" | "video/x-av1"
         | "image/jpeg" => write_visual_sample_entry(v, header, stream)?,
-        "audio/mpeg" | "audio/x-opus" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
+        "audio/mpeg" | "audio/x-opus" | "audio/x-flac" | "audio/x-alaw" | "audio/x-mulaw"
-            write_audio_sample_entry(v, header, stream)?
+        | "audio/x-adpcm" => write_audio_sample_entry(v, header, stream)?,
-        }
         "application/x-onvif-metadata" => write_xml_meta_data_sample_entry(v, header, stream)?,
         _ => unreachable!(),
     }
@@ -1079,6 +1077,7 @@ fn write_audio_sample_entry(
     let fourcc = match s.name().as_str() {
         "audio/mpeg" => b"mp4a",
         "audio/x-opus" => b"Opus",
+        "audio/x-flac" => b"fLaC",
         "audio/x-alaw" => b"alaw",
         "audio/x-mulaw" => b"ulaw",
         "audio/x-adpcm" => {
@@ -1097,6 +1096,10 @@ fn write_audio_sample_entry(
             let bitrate = s.get::<i32>("bitrate").context("no ADPCM bitrate field")?;
             (bitrate / 8000) as u16
         }
+        "audio/x-flac" => with_flac_metadata(&stream.caps, |streaminfo, _| {
+            1 + (u16::from_be_bytes([streaminfo[16], streaminfo[17]]) >> 4 & 0b11111)
+        })
+        .context("FLAC metadata error")?,
         _ => 16u16,
     };
 
@@ -1139,6 +1142,9 @@ fn write_audio_sample_entry(
             "audio/x-opus" => {
                 write_dops(v, &stream.caps)?;
             }
+            "audio/x-flac" => {
+                write_dfla(v, &stream.caps)?;
+            }
             "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
                 // Nothing to do here
             }
@@ -1333,6 +1339,35 @@ fn write_dops(v: &mut Vec<u8>, caps: &gst::Caps) -> Result<(), Error> {
     })
 }
 
+fn with_flac_metadata<R>(
+    caps: &gst::Caps,
+    cb: impl FnOnce(&[u8], &[gst::glib::SendValue]) -> R,
+) -> Result<R, Error> {
+    let caps = caps.structure(0).unwrap();
+    let header = caps.get::<gst::ArrayRef>("streamheader").unwrap();
+    let (streaminfo, remainder) = header.as_ref().split_first().unwrap();
+    let streaminfo = streaminfo.get::<&gst::BufferRef>().unwrap();
+    let streaminfo = streaminfo.map_readable().unwrap();
+    // 13 bytes for the Ogg/FLAC prefix and 38 for the streaminfo itself.
+    match <&[_; 13 + 38]>::try_from(streaminfo.as_slice()) {
+        Ok(i) if i.starts_with(b"\x7FFLAC\x01\x00") => Ok(cb(&i[13..], remainder)),
+        Ok(_) | Err(_) => bail!("Unknown streamheader format"),
+    }
+}
+
+fn write_dfla(v: &mut Vec<u8>, caps: &gst::Caps) -> Result<(), Error> {
+    write_full_box(v, b"dfLa", 0, 0, move |v| {
+        with_flac_metadata(caps, |streaminfo, remainder| {
+            v.extend(streaminfo);
+            for metadata in remainder {
+                let metadata = metadata.get::<&gst::BufferRef>().unwrap();
+                let metadata = metadata.map_readable().unwrap();
+                v.extend(&metadata[..]);
+            }
+        })
+    })
+}
+
 fn write_xml_meta_data_sample_entry(
     v: &mut Vec<u8>,
     _header: &super::Header,

mux/mp4/src/mp4mux/imp.rs

@@ -108,6 +108,8 @@ struct Stream {
     caps: gst::Caps,
     /// Whether this stream is intra-only and has frame reordering.
     delta_frames: super::DeltaFrames,
+    /// Whether this stream might have header frames without timestamps that should be ignored.
+    discard_header_buffers: bool,
 
     /// Already written out chunks with their samples for this stream
     chunks: Vec<super::Chunk>,
@@ -165,7 +167,12 @@ impl MP4Mux {
         buffer: &gst::BufferRef,
         sinkpad: &super::MP4MuxPad,
         delta_frames: super::DeltaFrames,
+        discard_headers: bool,
     ) -> Result<(), gst::FlowError> {
+        if discard_headers && buffer.flags().contains(gst::BufferFlags::HEADER) {
+            return Err(gst_base::AGGREGATOR_FLOW_NEED_DATA);
+        }
+
         if delta_frames.requires_dts() && buffer.dts().is_none() {
             gst::error!(CAT, obj: sinkpad, "Require DTS for video streams");
             return Err(gst::FlowError::Error);
@@ -188,6 +195,7 @@ impl MP4Mux {
         &self,
         sinkpad: &super::MP4MuxPad,
         delta_frames: super::DeltaFrames,
+        discard_headers: bool,
         pre_queue: &mut VecDeque<(gst::FormattedSegment<gst::ClockTime>, gst::Buffer)>,
         running_time_utc_time_mapping: &Option<(gst::Signed<gst::ClockTime>, gst::ClockTime)>,
     ) -> Result<Option<(gst::FormattedSegment<gst::ClockTime>, gst::Buffer)>, gst::FlowError> {
@@ -195,13 +203,10 @@ impl MP4Mux {
             return Ok(Some((segment.clone(), buffer.clone())));
         }
 
-        let mut buffer = match sinkpad.peek_buffer() {
+        let Some(mut buffer) = sinkpad.peek_buffer() else {
-            None => return Ok(None),
+            return Ok(None);
-            Some(buffer) => buffer,
         };
-
+        Self::check_buffer(&buffer, sinkpad, delta_frames, discard_headers)?;
-        Self::check_buffer(&buffer, sinkpad, delta_frames)?;
-
         let mut segment = match sinkpad.segment().downcast::<gst::ClockTime>().ok() {
             Some(segment) => segment,
             None => {
@@ -276,19 +281,20 @@ impl MP4Mux {
 
     fn pop_buffer(
         &self,
-        sinkpad: &super::MP4MuxPad,
+        stream: &mut Stream,
-        delta_frames: super::DeltaFrames,
-        pre_queue: &mut VecDeque<(gst::FormattedSegment<gst::ClockTime>, gst::Buffer)>,
-        running_time_utc_time_mapping: &mut Option<(gst::Signed<gst::ClockTime>, gst::ClockTime)>,
     ) -> Result<Option<(gst::FormattedSegment<gst::ClockTime>, gst::Buffer)>, gst::FlowError> {
+        let Stream {
+            sinkpad, pre_queue, ..
+        } = stream;
+
         // In ONVIF mode we need to get UTC times for each buffer and synchronize based on that.
         // Queue up to 6s of data to get the first UTC time and then backdate.
         if self.obj().class().as_ref().variant == super::Variant::ONVIF
-            && running_time_utc_time_mapping.is_none()
+            && stream.running_time_utc_time_mapping.is_none()
         {
             if let Some((last, first)) = Option::zip(pre_queue.back(), pre_queue.front()) {
                 // Existence of PTS/DTS checked below
-                let (last, first) = if delta_frames.requires_dts() {
+                let (last, first) = if stream.delta_frames.requires_dts() {
                     (
                         last.0.to_running_time_full(last.1.dts()).unwrap(),
                         first.0.to_running_time_full(first.1.dts()).unwrap(),
@@ -312,19 +318,20 @@ impl MP4Mux {
                 }
             }
 
-            let buffer = match sinkpad.pop_buffer() {
+            let Some(buffer) = sinkpad.pop_buffer() else {
-                None => {
                 if sinkpad.is_eos() {
                     gst::error!(CAT, obj: sinkpad, "Got no UTC time before EOS");
                     return Err(gst::FlowError::Error);
                 } else {
                     return Err(gst_base::AGGREGATOR_FLOW_NEED_DATA);
                 }
-                }
-                Some(buffer) => buffer,
             };
-
+            Self::check_buffer(
-            Self::check_buffer(&buffer, sinkpad, delta_frames)?;
+                &buffer,
+                sinkpad,
+                stream.delta_frames,
+                stream.discard_header_buffers,
+            )?;
 
             let segment = match sinkpad.segment().downcast::<gst::ClockTime>().ok() {
                 Some(segment) => segment,
@@ -350,7 +357,7 @@ impl MP4Mux {
             );
 
             let mapping = (running_time, utc_time);
-            *running_time_utc_time_mapping = Some(mapping);
+            stream.running_time_utc_time_mapping = Some(mapping);
 
             // Push the buffer onto the pre-queue and re-timestamp it and all other buffers
             // based on the mapping above.
@@ -391,7 +398,7 @@ impl MP4Mux {
             // Fall through below and pop the first buffer finally
         }
 
-        if let Some((segment, buffer)) = pre_queue.pop_front() {
+        if let Some((segment, buffer)) = stream.pre_queue.pop_front() {
             return Ok(Some((segment, buffer)));
         }
 
@@ -400,23 +407,26 @@ impl MP4Mux {
         //   for calculating the duration to the previous buffer, and then put into the pre-queue
         // - or this is the very first buffer and we just put it into the queue overselves above
         if self.obj().class().as_ref().variant == super::Variant::ONVIF {
-            if sinkpad.is_eos() {
+            if stream.sinkpad.is_eos() {
                 return Ok(None);
             }
             unreachable!();
         }
 
-        let buffer = match sinkpad.pop_buffer() {
+        let Some(buffer) = stream.sinkpad.pop_buffer() else {
-            None => return Ok(None),
+            return Ok(None);
-            Some(buffer) => buffer,
         };
+        Self::check_buffer(
+            &buffer,
+            &stream.sinkpad,
+            stream.delta_frames,
+            stream.discard_header_buffers,
+        )?;
 
-        Self::check_buffer(&buffer, sinkpad, delta_frames)?;
+        let segment = match stream.sinkpad.segment().downcast::<gst::ClockTime>().ok() {
-
-        let segment = match sinkpad.segment().downcast::<gst::ClockTime>().ok() {
             Some(segment) => segment,
             None => {
-                gst::error!(CAT, obj: sinkpad, "Got buffer before segment");
+                gst::error!(CAT, obj: stream.sinkpad, "Got buffer before segment");
                 return Err(gst::FlowError::Error);
             }
         };
@@ -442,6 +452,12 @@ impl MP4Mux {
                 Some(PendingBuffer {
                     duration: Some(_), ..
                 }) => return Ok(()),
+                Some(PendingBuffer { ref buffer, .. })
+                    if stream.discard_header_buffers
+                        && buffer.flags().contains(gst::BufferFlags::HEADER) =>
+                {
+                    return Err(gst_base::AGGREGATOR_FLOW_NEED_DATA);
+                }
                 Some(PendingBuffer {
                     timestamp,
                     pts,
@@ -449,13 +465,15 @@ impl MP4Mux {
                     ref mut duration,
                     ..
                 }) => {
-                    // Already have a pending buffer but no duration, so try to get that now
+                    let peek_outcome = self.peek_buffer(
-                    let (segment, buffer) = match self.peek_buffer(
                         &stream.sinkpad,
                         stream.delta_frames,
+                        stream.discard_header_buffers,
                         &mut stream.pre_queue,
                         &stream.running_time_utc_time_mapping,
-                    )? {
+                    )?;
+                    // Already have a pending buffer but no duration, so try to get that now
+                    let (segment, buffer) = match peek_outcome {
                         Some(res) => res,
                         None => {
                             if stream.sinkpad.is_eos() {
@@ -532,12 +550,7 @@ impl MP4Mux {
                 None => {
                     // Have no buffer queued at all yet
 
-                    let (segment, buffer) = match self.pop_buffer(
+                    let (segment, buffer) = match self.pop_buffer(stream)? {
-                        &stream.sinkpad,
-                        stream.delta_frames,
-                        &mut stream.pre_queue,
-                        &mut stream.running_time_utc_time_mapping,
-                    )? {
                         Some(res) => res,
                         None => {
                             if stream.sinkpad.is_eos() {
@@ -870,6 +883,7 @@ impl MP4Mux {
             let s = caps.structure(0).unwrap();
 
             let mut delta_frames = super::DeltaFrames::IntraOnly;
+            let mut discard_header_buffers = false;
             match s.name().as_str() {
                 "video/x-h264" | "video/x-h265" => {
                     if !s.has_field_with_type("codec_data", gst::Buffer::static_type()) {
@@ -913,6 +927,13 @@ impl MP4Mux {
                         return Err(gst::FlowError::NotNegotiated);
                     }
                 }
+                "audio/x-flac" => {
+                    discard_header_buffers = true;
+                    if let Err(e) = s.get::<gst::ArrayRef>("streamheader") {
+                        gst::error!(CAT, obj: pad, "Muxing FLAC into MP4 needs streamheader: {}", e);
+                        return Err(gst::FlowError::NotNegotiated);
+                    };
+                }
                 "audio/x-alaw" | "audio/x-mulaw" => (),
                 "audio/x-adpcm" => (),
                 "application/x-onvif-metadata" => (),
@@ -924,6 +945,7 @@ impl MP4Mux {
                 pre_queue: VecDeque::new(),
                 caps,
                 delta_frames,
+                discard_header_buffers,
                 chunks: Vec::new(),
                 pending_buffer: None,
                 queued_chunk_time: gst::ClockTime::ZERO,
@@ -1523,6 +1545,11 @@ impl ElementImpl for ISOMP4Mux {
                         .field("channels", gst::IntRange::new(1i32, 8))
                         .field("rate", gst::IntRange::new(1, i32::MAX))
                         .build(),
+                    gst::Structure::builder("audio/x-flac")
+                        .field("framed", true)
+                        .field("channels", gst::IntRange::<i32>::new(1, 8))
+                        .field("rate", gst::IntRange::<i32>::new(1, 10 * u16::MAX as i32))
+                        .build(),
                 ]
                 .into_iter()
                 .collect::<gst::Caps>(),

mux/mp4/tests/tests.rs

@@ -7,6 +7,8 @@
 // SPDX-License-Identifier: MPL-2.0
 //
 
+use std::path::Path;
+
 use gst::prelude::*;
 use gst_pbutils::prelude::*;
 
@@ -20,46 +22,26 @@ fn init() {
     });
 }
 
-#[test]
+struct Pipeline(gst::Pipeline);
-fn test_basic() {
+impl std::ops::Deref for Pipeline {
-    init();
-
-    struct Pipeline(gst::Pipeline);
-    impl std::ops::Deref for Pipeline {
     type Target = gst::Pipeline;
 
     fn deref(&self) -> &Self::Target {
         &self.0
     }
-    }
+}
-    impl Drop for Pipeline {
+impl Drop for Pipeline {
     fn drop(&mut self) {
         let _ = self.0.set_state(gst::State::Null);
     }
-    }
+}
 
-    let pipeline = match gst::parse::launch(
+impl Pipeline {
-        "videotestsrc num-buffers=99 ! x264enc ! mux. \
+    fn into_completion(self) {
-         audiotestsrc num-buffers=140 ! fdkaacenc ! mux. \
+        self.set_state(gst::State::Playing)
-         isomp4mux name=mux ! filesink name=sink \
-    ",
-    ) {
-        Ok(pipeline) => Pipeline(pipeline.downcast::<gst::Pipeline>().unwrap()),
-        Err(_) => return,
-    };
-
-    let dir = tempfile::TempDir::new().unwrap();
-    let mut location = dir.path().to_owned();
-    location.push("test.mp4");
-
-    let sink = pipeline.by_name("sink").unwrap();
-    sink.set_property("location", location.to_str().expect("Non-UTF8 filename"));
-
-    pipeline
-        .set_state(gst::State::Playing)
             .expect("Unable to set the pipeline to the `Playing` state");
 
-    for msg in pipeline.bus().unwrap().iter_timed(gst::ClockTime::NONE) {
+        for msg in self.bus().unwrap().iter_timed(gst::ClockTime::NONE) {
             use gst::MessageView;
 
             match msg.view() {
@@ -76,17 +58,42 @@ fn test_basic() {
             }
         }
 
-    pipeline
+        self.set_state(gst::State::Null)
-        .set_state(gst::State::Null)
             .expect("Unable to set the pipeline to the `Null` state");
+    }
+}
 
-    drop(pipeline);
+fn test_basic_with(video_enc: &str, audio_enc: &str, cb: impl FnOnce(&Path)) {
+    let Ok(pipeline) = gst::parse::launch(&format!(
+        "videotestsrc num-buffers=99 ! {video_enc} ! mux. \
+         audiotestsrc num-buffers=140 ! {audio_enc} ! mux. \
+         isomp4mux name=mux ! filesink name=sink"
+    )) else {
+        println!("could not build encoding pipeline");
+        return;
+    };
+    let pipeline = Pipeline(pipeline.downcast::<gst::Pipeline>().unwrap());
 
+    let dir = tempfile::TempDir::new().unwrap();
+    let mut location = dir.path().to_owned();
+    location.push("test.mp4");
+
+    let sink = pipeline.by_name("sink").unwrap();
+    sink.set_property("location", location.to_str().expect("Non-UTF8 filename"));
+    pipeline.into_completion();
+
+    cb(&location)
+}
+
+#[test]
+fn test_basic_x264_aac() {
+    init();
+    test_basic_with("x264enc", "fdkaacenc", |location| {
         let discoverer = gst_pbutils::Discoverer::new(gst::ClockTime::from_seconds(5))
             .expect("Failed to create discoverer");
         let info = discoverer
             .discover_uri(
-            url::Url::from_file_path(&location)
+                url::Url::from_file_path(location)
                     .expect("Failed to convert filename to URL")
                     .as_str(),
             )
@@ -123,4 +130,25 @@ fn test_basic() {
             caps.can_intersect(&gst::Caps::builder("video/x-h264").any_features().build()),
             "Unexpected video caps {caps:?}"
         );
+    })
+}
+
+#[test]
+fn test_roundtrip_vp9_flac() {
+    init();
+    test_basic_with("vp9enc ! vp9parse", "flacenc ! flacparse", |location| {
+        let Ok(pipeline) = gst::parse::launch(
+            "filesrc name=src ! qtdemux name=demux \
+             demux.audio_0 ! queue ! flacdec ! fakesink \
+             demux.video_0 ! queue ! vp9dec ! fakesink",
+        ) else {
+            panic!("could not build decoding pipeline")
+        };
+        let pipeline = Pipeline(pipeline.downcast::<gst::Pipeline>().unwrap());
+        pipeline
+            .by_name("src")
+            .unwrap()
+            .set_property("location", location.display().to_string());
+        pipeline.into_completion();
+    })
 }