Exercise codec output delay explicitly in the desktop fixture

tools/chatgui/CMakeLists.txt

@@ -537,7 +537,8 @@ if(CMAKE_SYSTEM_NAME STREQUAL "Linux")
     target_link_options(test_stream_view PRIVATE "-Wl,--wrap=ma_context_init" "-Wl,--wrap=ma_device_init"
         "-Wl,--wrap=audio_stream_frame_time")
     if(FFMPEG_FOUND)
-        target_link_options(test_stream_view PRIVATE "-Wl,--wrap=avcodec_open2")
+        target_link_options(test_stream_view PRIVATE "-Wl,--wrap=avcodec_open2" "-Wl,--wrap=avcodec_receive_frame"
+            "-Wl,--wrap=avcodec_free_context")
     endif()
 endif()
 add_test(NAME test_stream_view COMMAND test_stream_view)

tools/chatgui/tests/test_stream_view.cpp

@@ -10,6 +10,7 @@
 #include <atomic>
 #include <cmath>
 #include <chrono>
+#include <deque>
 
 extern "C" {
 #include "utun_instance.h"
@@ -29,12 +30,39 @@ extern "C" {
 static QList<QByteArray> snapshots;
 #if defined(TEST_NULL_AUDIO) && defined(HAVE_FFMPEG)
 static std::atomic<bool> delayedDecoder{false};
+static AVCodecContext* delayedContext = nullptr;
+static std::deque<AVFrame*> delayedFrames;
+static bool delayedDraining = false;
 extern "C" int __real_avcodec_open2(AVCodecContext*, const AVCodec*, AVDictionary**);
 extern "C" int __wrap_avcodec_open2(AVCodecContext* context, const AVCodec* codec, AVDictionary** options) {
-    if (delayedDecoder && av_codec_is_decoder(codec)) {
-        context->thread_count = 4; context->thread_type = FF_THREAD_FRAME;
+    int result = __real_avcodec_open2(context, codec, options);
+    if (!result && delayedDecoder && codec && av_codec_is_decoder(codec)) {
+        delayedContext = context; delayedDraining = false;
+        DEBUG_INFO(DEBUG_CATEGORY_STREAM, "fixture decoder name=%s holds four real decoded frames before output", codec->name);
     }
-    return __real_avcodec_open2(context, codec, options);
+    return result;
+}
+extern "C" int __real_avcodec_receive_frame(AVCodecContext*, AVFrame*);
+extern "C" int __wrap_avcodec_receive_frame(AVCodecContext* context, AVFrame* frame) {
+    if (context != delayedContext) return __real_avcodec_receive_frame(context, frame);
+    int result = __real_avcodec_receive_frame(context, frame);
+    if (!result) {
+        AVFrame* held = av_frame_clone(frame); if (!held) return AVERROR(ENOMEM);
+        delayedFrames.push_back(held); av_frame_unref(frame);
+    } else if (result == AVERROR_EOF) delayedDraining = true;
+    else if (result != AVERROR(EAGAIN)) return result;
+    if ((delayedDraining || delayedFrames.size() >= 4) && !delayedFrames.empty()) {
+        AVFrame* held = delayedFrames.front(); delayedFrames.pop_front(); av_frame_move_ref(frame, held); av_frame_free(&held); return 0;
+    }
+    return delayedDraining ? AVERROR_EOF : AVERROR(EAGAIN);
+}
+extern "C" void __real_avcodec_free_context(AVCodecContext**);
+extern "C" void __wrap_avcodec_free_context(AVCodecContext** context) {
+    if (context && *context == delayedContext) {
+        for (auto* frame : delayedFrames) av_frame_free(&frame);
+        delayedFrames.clear(); delayedContext = nullptr; delayedDraining = false;
+    }
+    __real_avcodec_free_context(context);
 }
 #endif
 #ifdef TEST_NULL_AUDIO