From 25d3fca079a19ee4905c5581c6e5c26ddd744ea7 Mon Sep 17 00:00:00 2001
From: Jim Broadus <jbroadus@gmail.com>
Date: Tue, 5 Jan 2021 15:38:08 -0800
Subject: [PATCH] gstengine: Fix output audio depth

When the decoder bin's src pad becomes available, check its caps
for a format. If the format is not S16LE, then apply F32LE caps
before the tee. This prevents the pipeline from negotiating
S16LE when the decoder produces other formats.
---
 src/engines/gstenginepipeline.cpp | 48 +++++++++++++++++++++++++++++--
 src/engines/gstenginepipeline.h   |  3 ++
 2 files changed, 49 insertions(+), 2 deletions(-)

diff --git a/src/engines/gstenginepipeline.cpp b/src/engines/gstenginepipeline.cpp
index 811e170c0..aac5dcc1a 100644
--- a/src/engines/gstenginepipeline.cpp
+++ b/src/engines/gstenginepipeline.cpp
@@ -850,20 +850,59 @@ void GstEnginePipeline::BufferingMessageReceived(GstMessage* msg) {
   }
 }
 
+QString GstEnginePipeline::GetAudioFormat(GstCaps* caps) {
+  const guint sz = gst_caps_get_size(caps);
+  for (int i = 0; i < sz; i++) {
+    GstStructure* s = gst_caps_get_structure(caps, i);
+    if (strcmp(gst_structure_get_name(s), "audio/x-raw") == 0) {
+      const gchar* fmt = gst_structure_get_string(s, "format");
+      if (fmt != nullptr) {
+        return QString::fromUtf8(fmt);
+      }
+    }
+  }
+  return "";
+}
+
 void GstEnginePipeline::NewPadCallback(GstElement*, GstPad* pad,
                                        gpointer self) {
   GstEnginePipeline* instance = reinterpret_cast<GstEnginePipeline*>(self);
   GstPad* const audiopad =
       gst_element_get_static_pad(instance->audiobin_, "sink");
 
-  // Link decodebin's sink pad to audiobin's src pad.
+  qLog(Debug) << "Decoder bin pad added:" << GST_PAD_NAME(pad);
+
+  // Make sure the audio bin isn't already linked to something.
   if (GST_PAD_IS_LINKED(audiopad)) {
     qLog(Warning) << instance->id()
                   << "audiopad is already linked, unlinking old pad";
     gst_pad_unlink(audiopad, GST_PAD_PEER(audiopad));
   }
 
-  gst_pad_link(pad, audiopad);
+  // See what the decoder bin wants to output.
+  GstCaps* caps = gst_pad_get_current_caps(pad);
+  if (caps) {
+    gchar* caps_str = gst_caps_to_string(caps);
+    qLog(Debug) << "Current caps:" << caps_str;
+    g_free(caps_str);
+
+    QString fmt = GetAudioFormat(caps);
+
+    // The output branch only handles F32LE and S16LE. If the source is S16LE,
+    // then use that throughout the pipeline. Otherwise, use F32LE.
+    if (fmt != "S16LE") {
+      GstCaps* new_caps = gst_caps_new_simple("audio/x-raw", "format",
+                                              G_TYPE_STRING, "F32LE", nullptr);
+      g_object_set(instance->capsfilter_, "caps", new_caps, nullptr);
+      gst_caps_unref(new_caps);
+    }
+    gst_caps_unref(caps);
+  }
+
+  // Link decodebin's sink pad to audiobin's src pad.
+  if (gst_pad_link(pad, audiopad) != GST_PAD_LINK_OK) {
+    qLog(Error) << "Failed to link decoder to audio bin.";
+  }
   gst_object_unref(audiopad);
 
   // Offset the timestamps on all the buffers coming out of the decodebin so
@@ -1100,6 +1139,11 @@ void GstEnginePipeline::TransitionToNext() {
 
   ignore_tags_ = true;
 
+  // Reset the caps filter
+  GstCaps* new_caps = gst_caps_new_any();
+  g_object_set(capsfilter_, "caps", new_caps, nullptr);
+  gst_caps_unref(new_caps);
+
   if (!ReplaceDecodeBin(next_.url_)) {
     qLog(Error) << "ReplaceDecodeBin failed with " << next_.url_;
     return;
diff --git a/src/engines/gstenginepipeline.h b/src/engines/gstenginepipeline.h
index 6a56764e2..039d54b03 100644
--- a/src/engines/gstenginepipeline.h
+++ b/src/engines/gstenginepipeline.h
@@ -168,6 +168,9 @@ class GstEnginePipeline : public QObject {
   // a src pad immediately and we can link it after everything's created.
   void MaybeLinkDecodeToAudio();
 
+  // Helper method to retrieve the audio format from a GstCaps object.
+  static QString GetAudioFormat(GstCaps* caps);
+
  private slots:
   void FaderTimelineFinished();