diff mbox series

[FFmpeg-devel,01/35] avdevice/dshow: implement option to use device video timestamps

Message ID 20210607230414.612-2-dcnieho@gmail.com
State Superseded, archived
Headers show
Series avdevice (mostly dshow) enhancements
Related show

Checks

Context Check Description
andriy/x86_make success Make finished
andriy/x86_make_fate success Make fate finished
andriy/PPC64_make success Make finished
andriy/PPC64_make_fate success Make fate finished

Commit Message

Diederick C. Niehorster June 7, 2021, 11:03 p.m. UTC
The dshow plugin ignores timestamps for video frames provided by the DirectShow device, instead using wallclock time, apparently because the implementer of this code had a device that provided unreliable timestamps. Me (and others) would like to use the device's timestamps. The new use_video_device_timestamps option for dshow device enables them to do so. Since the majority of video devices out there probably provide fine timestamps, this patch sets the default to using the device timestamps, which means best fidelity timestamps are used by default. Using the new option, the user can switch this off and revert to the old behavior, so a fall back remains available in case the device provides broken timestamps.

Closes: #8620

Signed-off-by: Diederick Niehorster <dcnieho@gmail.com>
---
 libavdevice/dshow.c         |  1 +
 libavdevice/dshow_capture.h |  1 +
 libavdevice/dshow_pin.c     | 15 ++++++++-------
 3 files changed, 10 insertions(+), 7 deletions(-)

Comments

Valerii Zapodovnikov June 7, 2021, 11:29 p.m. UTC | #1
Who knows what BS code "TODO figure out math. For now just drop them."
means? Is PTS of the mentioned there can be even theoretically valid or not?
Diederick C. Niehorster June 8, 2021, 5:43 a.m. UTC | #2
On Tue, Jun 8, 2021 at 1:29 AM Valerii Zapodovnikov <val.zapod.vz@gmail.com>
wrote:

> Who knows what BS code "TODO figure out math. For now just drop them."
> means? Is PTS of the mentioned there can be even theoretically valid or
> not?
>

No, I don't know. I am guessing it refers to wraparound? I also don't
understand why a negative time would lead to a large positive value, since
a signed integer is used. So i just left this intact. Without a test
case/device producing this, there's little that can be done i guess.
diff mbox series

Patch

diff --git a/libavdevice/dshow.c b/libavdevice/dshow.c
index 8d0a6fcc09..2e9f9ddf3f 100644
--- a/libavdevice/dshow.c
+++ b/libavdevice/dshow.c
@@ -1317,6 +1317,7 @@  static const AVOption options[] = {
     { "audio_device_save", "save audio capture filter device (and properties) to file", OFFSET(audio_filter_save_file), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, DEC },
     { "video_device_load", "load video capture filter device (and properties) from file", OFFSET(video_filter_load_file), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, DEC },
     { "video_device_save", "save video capture filter device (and properties) to file", OFFSET(video_filter_save_file), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, DEC },
+    { "use_video_device_timestamps", "use device instead of wallclock timestamps for video frames", OFFSET(use_video_device_timestamps), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, DEC },
     { NULL },
 };
 
diff --git a/libavdevice/dshow_capture.h b/libavdevice/dshow_capture.h
index 06ded2ba96..5a2691518c 100644
--- a/libavdevice/dshow_capture.h
+++ b/libavdevice/dshow_capture.h
@@ -312,6 +312,7 @@  struct dshow_ctx {
     char *audio_filter_save_file;
     char *video_filter_load_file;
     char *video_filter_save_file;
+    int   use_video_device_timestamps;
 
     IBaseFilter *device_filter[2];
     IPin        *device_pin[2];
diff --git a/libavdevice/dshow_pin.c b/libavdevice/dshow_pin.c
index 3dae405e65..ab0ead8041 100644
--- a/libavdevice/dshow_pin.c
+++ b/libavdevice/dshow_pin.c
@@ -309,12 +309,16 @@  long ff_dshow_meminputpin_Receive(DShowMemInputPin *this, IMediaSample *sample)
     if (!sample)
         return E_POINTER;
 
+    priv_data = pin->filter->priv_data;
+    s = priv_data;
+    ctx = s->priv_data;
+
     IMediaSample_GetTime(sample, &orig_curtime, &dummy);
     orig_curtime += pin->filter->start_time;
     IReferenceClock_GetTime(clock, &graphtime);
-    if (devtype == VideoDevice) {
-        /* PTS from video devices is unreliable. */
-        IReferenceClock_GetTime(clock, &curtime);
+    if (devtype == VideoDevice && !ctx->use_video_device_timestamps) {
+            /* PTS from video devices is unreliable. */
+            IReferenceClock_GetTime(clock, &curtime);
     } else {
         IMediaSample_GetTime(sample, &curtime, &dummy);
         if(curtime > 400000000000000000LL) {
@@ -322,7 +326,7 @@  long ff_dshow_meminputpin_Receive(DShowMemInputPin *this, IMediaSample *sample)
                like 437650244077016960 which FFmpeg doesn't like.
                TODO figure out math. For now just drop them. */
             av_log(NULL, AV_LOG_DEBUG,
-                "dshow dropping initial (or ending) audio frame with odd PTS too high %"PRId64"\n", curtime);
+                "dshow dropping initial (or ending) frame with odd PTS too high %"PRId64"\n", curtime);
             return S_OK;
         }
         curtime += pin->filter->start_time;
@@ -330,9 +334,6 @@  long ff_dshow_meminputpin_Receive(DShowMemInputPin *this, IMediaSample *sample)
 
     buf_size = IMediaSample_GetActualDataLength(sample);
     IMediaSample_GetPointer(sample, &buf);
-    priv_data = pin->filter->priv_data;
-    s = priv_data;
-    ctx = s->priv_data;
     index = pin->filter->stream_index;
 
     av_log(NULL, AV_LOG_VERBOSE, "dshow passing through packet of type %s size %8d "