gst-plugins-rs/video/closedcaption/tests/tttocea608.rs

389 lines
11 KiB
Rust
Raw Normal View History

// Copyright (C) 2020 Mathieu Duponchelle <mathieu@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 Street, Suite 500,
// Boston, MA 02110-1335, USA.
2020-11-22 17:21:45 +00:00
use pretty_assertions::assert_eq;
fn init() {
use std::sync::Once;
static INIT: Once = Once::new();
INIT.call_once(|| {
gst::init().unwrap();
gstrsclosedcaption::plugin_register_static().unwrap();
});
}
fn new_timed_buffer<T: AsRef<[u8]> + Send + 'static>(
slice: T,
timestamp: gst::ClockTime,
duration: gst::ClockTime,
) -> gst::buffer::Buffer {
let mut buf = gst::Buffer::from_slice(slice);
let buf_ref = buf.get_mut().unwrap();
buf_ref.set_pts(timestamp);
buf_ref.set_duration(duration);
buf
}
#[test]
fn test_non_timed_buffer() {
init();
let mut h = gst_check::Harness::new_parse("tttocea608 mode=pop-on");
h.set_src_caps_str("text/x-raw");
let inbuf = gst::Buffer::from_slice(&"Hello");
assert_eq!(h.push(inbuf), Err(gst::FlowError::Error));
}
/* Check translation of a simple string */
#[test]
fn test_one_timed_buffer_and_eos() {
init();
let mut h = gst_check::Harness::new_parse("tttocea608 mode=pop-on");
h.set_src_caps_str("text/x-raw");
while h.events_in_queue() != 0 {
let _event = h.pull_event().unwrap();
}
let inbuf = new_timed_buffer(&"Hello", gst::SECOND, gst::SECOND);
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
loop {
let outbuf = h.pull().unwrap();
if outbuf.get_pts() + outbuf.get_duration() >= gst::SECOND {
break;
}
let data = outbuf.map_readable().unwrap();
assert_eq!(&*data, &[0x80, 0x80]);
}
let expected: [(gst::ClockTime, gst::ClockTime, [u8; 2usize]); 7] = [
(1_000_000_000.into(), 33_333_333.into(), [0x94, 0x20]), /* resume_caption_loading */
(1_033_333_333.into(), 33_333_334.into(), [0x94, 0xae]), /* erase_non_displayed_memory */
(1_066_666_667.into(), 33_333_333.into(), [0x94, 0x70]), /* preamble */
(1_100_000_000.into(), 33_333_333.into(), [0xc8, 0xe5]), /* H e */
(1_133_333_333.into(), 33_333_334.into(), [0xec, 0xec]), /* l l */
(1_166_666_667.into(), 33_333_333.into(), [0xef, 0x80]), /* o, nil */
(1_200_000_000.into(), 33_333_333.into(), [0x94, 0x2f]), /* end_of_caption */
];
for (i, e) in expected.iter().enumerate() {
let outbuf = h.try_pull().unwrap();
assert_eq!(
e.0,
outbuf.get_pts(),
"Unexpected PTS for {}th buffer",
i + 1
);
assert_eq!(
e.1,
outbuf.get_duration(),
"Unexpected duration for {}th buffer",
i + 1
);
let data = outbuf.map_readable().unwrap();
assert_eq!(e.2, &*data);
}
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
assert_eq!(h.buffers_in_queue(), 0);
h.push_event(gst::event::Eos::new());
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
/* Check that we do receive an erase_display */
loop {
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let outbuf = h.try_pull().unwrap();
let data = outbuf.map_readable().unwrap();
if outbuf.get_pts() == 2_200_000_000.into() {
assert_eq!(&*data, &[0x94, 0x2c]);
break;
} else {
assert_eq!(&*data, &[0x80, 0x80]);
}
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
}
assert_eq!(h.events_in_queue() == 1, true);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let event = h.pull_event().unwrap();
assert_eq!(event.get_type(), gst::EventType::Eos);
}
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
/* Here we test that the erase_display_memory control code
* gets inserted at the correct moment, when there's enough
* of an interval between two buffers
*/
#[test]
fn test_erase_display_memory_non_spliced() {
init();
let mut h = gst_check::Harness::new_parse("tttocea608 mode=pop-on");
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
h.set_src_caps_str("text/x-raw");
while h.events_in_queue() != 0 {
let _event = h.pull_event().unwrap();
}
let inbuf = new_timed_buffer(&"Hello", 1_000_000_000.into(), gst::SECOND);
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
let inbuf = new_timed_buffer(&"World", 3_000_000_000.into(), gst::SECOND);
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
let mut erase_display_buffers = 0;
while h.buffers_in_queue() > 0 {
let outbuf = h.pull().unwrap();
if outbuf.get_pts() == 2_200_000_000.into() {
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let data = outbuf.map_readable().unwrap();
assert_eq!(&*data, &[0x94, 0x2c]);
erase_display_buffers += 1;
}
}
assert_eq!(erase_display_buffers, 1);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
}
/* Here we test that the erase_display_memory control code
* gets inserted before the following pop-on captions
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
* when there's not enough of an interval between them.
*/
#[test]
fn test_erase_display_memory_spliced() {
init();
let mut h = gst_check::Harness::new_parse("tttocea608 mode=pop-on");
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
h.set_src_caps_str("text/x-raw");
while h.events_in_queue() != 0 {
let _event = h.pull_event().unwrap();
}
let inbuf = new_timed_buffer(&"Hello", 1_000_000_000.into(), gst::SECOND);
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
let inbuf = new_timed_buffer(&"World", 2_000_000_000.into(), gst::SECOND);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
let mut erase_display_buffers = 0;
let mut prev_pts: gst::ClockTime = 0.into();
while h.buffers_in_queue() > 0 {
let outbuf = h.pull().unwrap();
/* Check that our timestamps are strictly ascending */
assert!(outbuf.get_pts() >= prev_pts);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
if outbuf.get_pts() == 2_000_000_000.into() {
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let data = outbuf.map_readable().unwrap();
assert_eq!(&*data, &[0x94, 0x2c]);
erase_display_buffers += 1;
}
prev_pts = outbuf.get_pts();
}
assert_eq!(erase_display_buffers, 1);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
}
/* Here we verify that the element outputs a continuous stream
* with padding buffers
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
*/
#[test]
fn test_output_gaps() {
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
init();
let mut h = gst_check::Harness::new_parse("tttocea608 mode=pop-on");
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
h.set_src_caps_str("text/x-raw");
while h.events_in_queue() != 0 {
let _event = h.pull_event().unwrap();
}
let inbuf = new_timed_buffer(&"Hello", 1_000_000_000.into(), gst::SECOND);
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
let inbuf = new_timed_buffer(&"World", 3_000_000_000.into(), gst::SECOND);
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
h.push_event(gst::event::Eos::new());
/* Padding */
loop {
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let outbuf = h.pull().unwrap();
if outbuf.get_pts() + outbuf.get_duration() >= gst::SECOND {
break;
}
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let data = outbuf.map_readable().unwrap();
assert_eq!(&*data, &[0x80, 0x80]);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
}
/* Hello */
loop {
let outbuf = h.pull().unwrap();
if outbuf.get_pts() + outbuf.get_duration() >= 1_233_333_333.into() {
break;
}
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let data = outbuf.map_readable().unwrap();
assert_ne!(&*data, &[0x80, 0x80]);
}
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
/* Padding */
loop {
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let outbuf = h.pull().unwrap();
if outbuf.get_pts() + outbuf.get_duration() >= 3_000_000_000.into() {
break;
}
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let data = outbuf.map_readable().unwrap();
if outbuf.get_pts() == 2_200_000_000.into() {
/* Erase display one second after Hello */
assert_eq!(&*data, &[0x94, 0x2C]);
} else {
assert_eq!(&*data, &[0x80, 0x80]);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
}
}
/* World */
loop {
let outbuf = h.pull().unwrap();
if outbuf.get_pts() + outbuf.get_duration() >= 3_233_333_333.into() {
break;
}
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let data = outbuf.map_readable().unwrap();
assert_ne!(&*data, &[0x80, 0x80]);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
}
assert_eq!(h.events_in_queue(), 1);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
let event = h.pull_event().unwrap();
assert_eq!(event.get_type(), gst::EventType::Eos);
tttocea608: refactor to fit more scenarios - Report a latency: By design, tttocea608 will output buffers in the "past" when receiving an input buffer: we want the second to last buffer in the buffer list that we output to have the same pts as the input buffer, as it contains the end_of_caption control code which determines when the current closed caption actually gets displayed in pop_on mode. The previous buffers have timestamps decreasing as a function of the framerate, for up to potentially 74 byte pairs (the breakdown is detailed in a comment). The element thus has to report a latency, at 30 frames per second it represents around 2.5 seconds. - Refactor timestamping: Stop using a frame duration, but rather base our timestamps on a scaled frame index. This is to avoid rounding errors, and allow for exactly one byte pair per buffer if the proper framerate is set on the closed caption branch, and the video branch has perfect timestamps, eg videorate. In practice, that one byte pair per frame requirement should only matter for line 21 encoding, but we have to think about this use case too. - Splice in erase_display_memory: When there is a gap between the end of a buffer and the start of the next one, we want to erase the display memory (this is unnecessary otherwise, as the end_of_caption control code will in effect ensure that the display is erased when the new caption is displayed). The previous implementation only supported this imperfectly, as it could cause timestamps to go backwards. - Output last erase_display_memory: The previous implementation was missing the final erase_display_memory on EOS - Output gaps - Write more tests Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-rs/-/merge_requests/314>
2020-04-21 22:23:28 +00:00
}
#[test]
fn test_one_timed_buffer_and_eos_roll_up2() {
init();
let mut h = gst_check::Harness::new_parse("tttocea608 mode=roll-up2");
h.set_src_caps_str("text/x-raw");
while h.events_in_queue() != 0 {
let _event = h.pull_event().unwrap();
}
let inbuf = new_timed_buffer(&"Hello", gst::SECOND, gst::SECOND);
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
let inbuf = new_timed_buffer(&"World", 2 * gst::SECOND, 1.into());
assert_eq!(h.push(inbuf), Ok(gst::FlowSuccess::Ok));
/* Padding */
loop {
let outbuf = h.pull().unwrap();
if outbuf.get_pts() + outbuf.get_duration() >= gst::SECOND {
break;
}
let data = outbuf.map_readable().unwrap();
assert_eq!(&*data, &[0x80, 0x80]);
}
let expected: [(gst::ClockTime, gst::ClockTime, [u8; 2usize]); 5] = [
(1_000_000_000.into(), 33_333_333.into(), [0x94, 0x25]), /* roll_up_2 */
(1_033_333_333.into(), 33_333_334.into(), [0x94, 0x70]), /* preamble */
(1_066_666_667.into(), 33_333_333.into(), [0xc8, 0xe5]), /* H e */
(1_100_000_000.into(), 33_333_333.into(), [0xec, 0xec]), /* l l */
(1_133_333_333.into(), 33_333_334.into(), [0xef, 0x80]), /* o nil */
];
for (i, e) in expected.iter().enumerate() {
let outbuf = h.try_pull().unwrap();
assert_eq!(
e.0,
outbuf.get_pts(),
"Unexpected PTS for {}th buffer",
i + 1
);
assert_eq!(
e.1,
outbuf.get_duration(),
"Unexpected duration for {}th buffer",
i + 1
);
let data = outbuf.map_readable().unwrap();
assert_eq!(e.2, &*data);
}
/* Padding */
loop {
let outbuf = h.pull().unwrap();
if outbuf.get_pts() + outbuf.get_duration() >= 2 * gst::SECOND {
break;
}
let data = outbuf.map_readable().unwrap();
assert_eq!(&*data, &[0x80, 0x80]);
}
let expected: [(gst::ClockTime, gst::ClockTime, [u8; 2usize]); 3] = [
(2_000_000_000.into(), 0.into(), [0x20, 0x57]), /* SPACE W */
(2_000_000_000.into(), 0.into(), [0xef, 0xf2]), /* o r */
(2_000_000_000.into(), 0.into(), [0xec, 0x64]), /* l d */
];
for (i, e) in expected.iter().enumerate() {
let outbuf = h.try_pull().unwrap();
assert_eq!(
e.0,
outbuf.get_pts(),
"Unexpected PTS for {}th buffer",
i + 1
);
assert_eq!(
e.1,
outbuf.get_duration(),
"Unexpected duration for {}th buffer",
i + 1
);
let data = outbuf.map_readable().unwrap();
assert_eq!(e.2, &*data);
}
assert_eq!(h.buffers_in_queue(), 0);
h.push_event(gst::event::Eos::new());
assert_eq!(h.events_in_queue(), 1);
let event = h.pull_event().unwrap();
assert_eq!(event.get_type(), gst::EventType::Eos);
}