fmp4mux: Add support for ONVIF timed XML metadata

generic/fmp4/src/fmp4mux/boxes.rs

@@ -644,10 +644,11 @@ fn write_hdlr(
 
     let s = caps.structure(0).unwrap();
     let (handler_type, name) = match s.name() {
-        "video/x-h264" | "video/x-h265" | "image/jpeg" => (b"vide", b"VideoHandler\0"),
+        "video/x-h264" | "video/x-h265" | "image/jpeg" => (b"vide", b"VideoHandler\0".as_slice()),
         "audio/mpeg" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
-            (b"soun", b"SoundHandler\0")
+            (b"soun", b"SoundHandler\0".as_slice())
         }
+        "application/x-onvif-metadata" => (b"meta", b"MetadataHandler\0".as_slice()),
         _ => unreachable!(),
     };
 
@@ -680,6 +681,11 @@ fn write_minf(
                 write_smhd(v, cfg)
             })?
         }
+        "application/x-onvif-metadata" => {
+            write_full_box(v, b"nmhd", FULL_BOX_VERSION_0, FULL_BOX_FLAGS_NONE, |_v| {
+                Ok(())
+            })?
+        }
         _ => unreachable!(),
     }
 
@@ -785,6 +791,7 @@ fn write_stsd(
         "audio/mpeg" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" => {
             write_audio_sample_entry(v, cfg, caps)?
         }
+        "application/x-onvif-metadata" => write_xml_meta_data_sample_entry(v, cfg, caps)?,
         _ => unreachable!(),
     }
 
@@ -1246,6 +1253,34 @@ fn write_esds_aac(v: &mut Vec<u8>, codec_data: &[u8]) -> Result<(), Error> {
     )
 }
 
+fn write_xml_meta_data_sample_entry(
+    v: &mut Vec<u8>,
+    _cfg: &super::HeaderConfiguration,
+    caps: &gst::CapsRef,
+) -> Result<(), Error> {
+    let s = caps.structure(0).unwrap();
+    let namespace = match s.name() {
+        "application/x-onvif-metadata" => b"http://www.onvif.org/ver10/schema",
+        _ => unreachable!(),
+    };
+
+    write_sample_entry_box(v, b"metx", move |v| {
+        // content_encoding, empty string
+        v.push(0);
+
+        // namespace
+        v.extend_from_slice(namespace);
+        v.push(0);
+
+        // schema_location, empty string list
+        v.push(0);
+
+        Ok(())
+    })?;
+
+    Ok(())
+}
+
 fn write_stts(v: &mut Vec<u8>, _cfg: &super::HeaderConfiguration) -> Result<(), Error> {
     // Entry count
     v.extend(0u32.to_be_bytes());
@@ -1681,7 +1716,12 @@ fn write_traf(
     let check_dts = matches!(s.name(), "video/x-h264" | "video/x-h265");
     let intra_only = matches!(
         s.name(),
-        "audio/mpeg" | "audio/x-alaw" | "audio/x-mulaw" | "audio/x-adpcm" | "image/jpeg"
+        "audio/mpeg"
+            | "audio/x-alaw"
+            | "audio/x-mulaw"
+            | "audio/x-adpcm"
+            | "image/jpeg"
+            | "application/x-onvif-metadata"
     );
 
     // Analyze all buffers to know what values can be put into the tfhd for all samples and what

generic/fmp4/src/fmp4mux/imp.rs

@@ -860,6 +860,9 @@ impl FMP4Mux {
                 "audio/x-adpcm" => {
                     intra_only = true;
                 }
+                "application/x-onvif-metadata" => {
+                    intra_only = true;
+                }
                 _ => unreachable!(),
             }
 
@@ -879,32 +882,30 @@ impl FMP4Mux {
             return Err(gst::FlowError::Error);
         }
 
-        // Sort video streams first and then audio streams, and each group by pad name.
+        // Sort video streams first and then audio streams and then metadata streams, and each group by pad name.
         state.streams.sort_by(|a, b| {
-            let stream_type_of_caps = |caps: &gst::CapsRef| {
+            let order_of_caps = |caps: &gst::CapsRef| {
                 let s = caps.structure(0).unwrap();
 
                 if s.name().starts_with("video/") {
-                    gst::StreamType::VIDEO
+                    0
                 } else if s.name().starts_with("audio/") {
-                    gst::StreamType::AUDIO
+                    1
+                } else if s.name().starts_with("application/x-onvif-metadata") {
+                    2
                 } else {
                     unimplemented!();
                 }
             };
 
-            let st_a = stream_type_of_caps(&a.caps);
-            let st_b = stream_type_of_caps(&b.caps);
+            let st_a = order_of_caps(&a.caps);
+            let st_b = order_of_caps(&b.caps);
 
             if st_a == st_b {
                 return a.sinkpad.name().cmp(&b.sinkpad.name());
             }
 
-            if st_a == gst::StreamType::VIDEO {
-                std::cmp::Ordering::Less
-            } else {
-                std::cmp::Ordering::Greater
-            }
+            st_a.cmp(&st_b)
         });
 
         Ok(())
@@ -1963,6 +1964,9 @@ impl ElementImpl for ONVIFFMP4Mux {
                         .field("rate", 8000i32)
                         .field("bitrate", gst::List::new([16000i32, 24000, 32000, 40000]))
                         .build(),
+                    gst::Structure::builder("application/x-onvif-metadata")
+                        .field("encoding", "utf8")
+                        .build(),
                 ]
                 .into_iter()
                 .collect::<gst::Caps>(),