mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-11-18 15:51:11 +00:00
275 lines
8.1 KiB
C
275 lines
8.1 KiB
C
/* GStreamer interactive test for accurate seeking
|
|
* Copyright (C) 2014 Tim-Philipp Müller <tim centricular com>
|
|
*
|
|
* This library is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU Library General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2 of the License, or (at your option) any later version.
|
|
*
|
|
* This library is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* Library General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Library General Public
|
|
* License along with this library; if not, write to the
|
|
* Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
|
|
* Boston, MA 02110-1301, USA.
|
|
*
|
|
* Based on python script by Kibeom Kim <kkb110@gmail.com>
|
|
*/
|
|
#ifdef HAVE_CONFIG_H
|
|
# include "config.h"
|
|
#endif
|
|
|
|
#define _GNU_SOURCE /* for memmem */
|
|
#include <string.h>
|
|
|
|
#include <gst/gst.h>
|
|
#include <gst/base/base.h>
|
|
#include <gst/audio/audio.h>
|
|
#include <gst/app/app.h>
|
|
|
|
#define SAMPLE_FREQ 44100
|
|
|
|
static GstClockTime
|
|
sample_to_nanotime (guint sample)
|
|
{
|
|
return (guint64) ((1.0 * sample * GST_SECOND / SAMPLE_FREQ) + 0.5);
|
|
}
|
|
|
|
static guint
|
|
nanotime_to_sample (GstClockTime nanotime)
|
|
{
|
|
return gst_util_uint64_scale_round (nanotime, SAMPLE_FREQ, GST_SECOND);
|
|
}
|
|
|
|
static GstBuffer *
|
|
generate_test_data (guint N)
|
|
{
|
|
gint16 *left, *right, *stereo;
|
|
guint largeN, i, j;
|
|
|
|
/* 32767 = (2 ** 15) - 1 */
|
|
/* 32768 = (2 ** 15) */
|
|
largeN = ((N + 32767) / 32768) * 32768;
|
|
left = g_new0 (gint16, largeN);
|
|
right = g_new0 (gint16, largeN);
|
|
stereo = g_new0 (gint16, 2 * largeN);
|
|
|
|
for (i = 0; i < (largeN / 32768); ++i) {
|
|
gint c = 0;
|
|
|
|
for (j = i * 32768; j < ((i + 1) * 32768); ++j) {
|
|
left[j] = i;
|
|
|
|
if (i % 2 == 0) {
|
|
right[j] = c;
|
|
} else {
|
|
right[j] = 32767 - c;
|
|
}
|
|
++c;
|
|
}
|
|
}
|
|
|
|
/* could just fill stereo directly from the start, but keeping original code for now */
|
|
for (i = 0; i < largeN; ++i) {
|
|
stereo[(2 * i) + 0] = left[i];
|
|
stereo[(2 * i) + 1] = right[i];
|
|
}
|
|
g_free (left);
|
|
g_free (right);
|
|
|
|
return gst_buffer_new_wrapped (stereo, 2 * largeN * sizeof (gint16));
|
|
}
|
|
|
|
static void
|
|
generate_test_sound (const gchar * fn, const gchar * launch_string,
|
|
guint num_samples)
|
|
{
|
|
GstElement *pipeline, *src, *parse, *enc_bin, *sink;
|
|
GstFlowReturn flow;
|
|
GstMessage *msg;
|
|
GstBuffer *buf;
|
|
GstCaps *caps;
|
|
|
|
pipeline = gst_pipeline_new (NULL);
|
|
|
|
src = gst_element_factory_make ("appsrc", NULL);
|
|
|
|
caps = gst_caps_new_simple ("audio/x-raw",
|
|
"format", G_TYPE_STRING, GST_AUDIO_NE (S16),
|
|
"rate", G_TYPE_INT, SAMPLE_FREQ, "channels", G_TYPE_INT, 2,
|
|
"layout", G_TYPE_STRING, "interleaved",
|
|
"channel-mask", GST_TYPE_BITMASK, (guint64) 3, NULL);
|
|
g_object_set (src, "caps", caps, "format", GST_FORMAT_TIME, NULL);
|
|
gst_base_src_set_format (GST_BASE_SRC (src), GST_FORMAT_TIME);
|
|
gst_caps_unref (caps);
|
|
|
|
/* audioparse to put proper timestamps on buffers for us, without which
|
|
* vorbisenc in particular is unhappy (or oggmux, rather) */
|
|
parse = gst_element_factory_make ("audioparse", NULL);
|
|
if (parse != NULL) {
|
|
g_object_set (parse, "use-sink-caps", TRUE, NULL);
|
|
} else {
|
|
parse = gst_element_factory_make ("identity", NULL);
|
|
g_warning ("audioparse element not available, vorbis/ogg might not work\n");
|
|
}
|
|
|
|
enc_bin = gst_parse_bin_from_description (launch_string, TRUE, NULL);
|
|
|
|
sink = gst_element_factory_make ("filesink", NULL);
|
|
g_object_set (sink, "location", fn, NULL);
|
|
|
|
gst_bin_add_many (GST_BIN (pipeline), src, parse, enc_bin, sink, NULL);
|
|
|
|
gst_element_link_many (src, parse, enc_bin, sink, NULL);
|
|
|
|
gst_element_set_state (pipeline, GST_STATE_PLAYING);
|
|
|
|
buf = generate_test_data (num_samples);
|
|
flow = gst_app_src_push_buffer (GST_APP_SRC (src), buf);
|
|
g_assert (flow == GST_FLOW_OK);
|
|
|
|
gst_app_src_end_of_stream (GST_APP_SRC (src));
|
|
|
|
/*g_print ("generating test sound %s, waiting for EOS..\n", fn); */
|
|
|
|
msg = gst_bus_timed_pop_filtered (GST_ELEMENT_BUS (pipeline),
|
|
GST_CLOCK_TIME_NONE, GST_MESSAGE_EOS | GST_MESSAGE_ERROR);
|
|
|
|
g_assert (GST_MESSAGE_TYPE (msg) == GST_MESSAGE_EOS);
|
|
gst_message_unref (msg);
|
|
|
|
gst_element_set_state (pipeline, GST_STATE_NULL);
|
|
gst_object_unref (pipeline);
|
|
|
|
/* g_print ("Done %s\n", fn); */
|
|
}
|
|
|
|
static void
|
|
test_seek_FORMAT_TIME_by_sample (const gchar * fn, GList * seek_positions)
|
|
{
|
|
GstElement *pipeline, *src, *sink;
|
|
GstAdapter *adapter;
|
|
GstSample *sample;
|
|
GstCaps *caps;
|
|
gconstpointer answer;
|
|
guint answer_size;
|
|
|
|
pipeline = gst_parse_launch ("filesrc name=src ! decodebin ! "
|
|
"audioconvert dithering=0 ! appsink name=sink", NULL);
|
|
|
|
src = gst_bin_get_by_name (GST_BIN (pipeline), "src");
|
|
g_object_set (src, "location", fn, NULL);
|
|
gst_object_unref (src);
|
|
|
|
sink = gst_bin_get_by_name (GST_BIN (pipeline), "sink");
|
|
caps = gst_caps_new_simple ("audio/x-raw",
|
|
"format", G_TYPE_STRING, GST_AUDIO_NE (S16),
|
|
"rate", G_TYPE_INT, SAMPLE_FREQ, "channels", G_TYPE_INT, 2, NULL);
|
|
g_object_set (sink, "caps", caps, "sync", FALSE, NULL);
|
|
gst_caps_unref (caps);
|
|
|
|
gst_element_set_state (pipeline, GST_STATE_PLAYING);
|
|
|
|
/* wait for preroll, so we can seek */
|
|
gst_bus_timed_pop_filtered (GST_ELEMENT_BUS (pipeline), GST_CLOCK_TIME_NONE,
|
|
GST_MESSAGE_ASYNC_DONE);
|
|
|
|
/* first, read entire file to end */
|
|
adapter = gst_adapter_new ();
|
|
while ((sample = gst_app_sink_pull_sample (GST_APP_SINK (sink)))) {
|
|
gst_adapter_push (adapter, gst_buffer_ref (gst_sample_get_buffer (sample)));
|
|
gst_sample_unref (sample);
|
|
}
|
|
answer_size = gst_adapter_available (adapter);
|
|
answer = gst_adapter_map (adapter, answer_size);
|
|
/* g_print ("%s: read %u bytes\n", fn, answer_size); */
|
|
|
|
g_print ("%10s\t%10s\t%10s\n", "requested", "sample per ts", "actual(data)");
|
|
|
|
while (seek_positions != NULL) {
|
|
gconstpointer found;
|
|
GstMapInfo map;
|
|
GstBuffer *buf;
|
|
gboolean ret;
|
|
guint actual_position, buffer_timestamp_position;
|
|
guint seek_sample;
|
|
|
|
seek_sample = GPOINTER_TO_UINT (seek_positions->data);
|
|
|
|
ret = gst_element_seek_simple (pipeline, GST_FORMAT_TIME,
|
|
GST_SEEK_FLAG_FLUSH | GST_SEEK_FLAG_ACCURATE,
|
|
sample_to_nanotime (seek_sample));
|
|
|
|
g_assert (ret);
|
|
|
|
sample = gst_app_sink_pull_sample (GST_APP_SINK (sink));
|
|
|
|
buf = gst_sample_get_buffer (sample);
|
|
gst_buffer_map (buf, &map, GST_MAP_READ);
|
|
found = memmem (answer, answer_size, map.data, map.size);
|
|
gst_buffer_unmap (buf, &map);
|
|
|
|
g_assert (found != NULL);
|
|
actual_position = ((goffset) ((guint8 *) found - (guint8 *) answer)) / 4;
|
|
buffer_timestamp_position = nanotime_to_sample (GST_BUFFER_PTS (buf));
|
|
g_print ("%10u\t%10u\t%10u\n", seek_sample, buffer_timestamp_position,
|
|
actual_position);
|
|
gst_sample_unref (sample);
|
|
|
|
seek_positions = seek_positions->next;
|
|
}
|
|
|
|
gst_element_set_state (pipeline, GST_STATE_NULL);
|
|
gst_object_unref (sink);
|
|
gst_object_unref (pipeline);
|
|
g_object_unref (adapter);
|
|
}
|
|
|
|
static GList *
|
|
create_test_samples (guint from, guint to, guint step)
|
|
{
|
|
GQueue q = G_QUEUE_INIT;
|
|
guint i;
|
|
|
|
for (i = from; i < to; i += step)
|
|
g_queue_push_tail (&q, GUINT_TO_POINTER (i));
|
|
|
|
return q.head;
|
|
}
|
|
|
|
#define SECS 10
|
|
|
|
int
|
|
main (int argc, char **argv)
|
|
{
|
|
GList *test_samples;
|
|
|
|
gst_init (&argc, &argv);
|
|
|
|
test_samples = create_test_samples (SAMPLE_FREQ, SAMPLE_FREQ * 2, 5000);
|
|
|
|
g_print ("\nwav:\n");
|
|
generate_test_sound ("test.wav", "wavenc", SAMPLE_FREQ * SECS);
|
|
test_seek_FORMAT_TIME_by_sample ("test.wav", test_samples);
|
|
|
|
g_print ("\nflac:\n");
|
|
generate_test_sound ("test.flac", "flacenc", SAMPLE_FREQ * SECS);
|
|
test_seek_FORMAT_TIME_by_sample ("test.flac", test_samples);
|
|
|
|
g_print ("\nogg:\n");
|
|
generate_test_sound ("test.ogg",
|
|
"audioconvert dithering=0 ! vorbisenc quality=1 ! oggmux",
|
|
SAMPLE_FREQ * SECS);
|
|
test_seek_FORMAT_TIME_by_sample ("test.ogg", test_samples);
|
|
|
|
g_print ("\nmp3:\n");
|
|
generate_test_sound ("test.mp3", "lamemp3enc bitrate=320",
|
|
SAMPLE_FREQ * SECS);
|
|
test_seek_FORMAT_TIME_by_sample ("test.mp3", test_samples);
|
|
|
|
g_list_free (test_samples);
|
|
return 0;
|
|
}
|