Files
terebi/Checks/cast-media.cpp
T

326 lines
23 KiB
C++

#include "cast_media.h"
extern "C" {
#include <libavcodec/avcodec.h>
#include <libavcodec/bsf.h>
#include <libavformat/avformat.h>
#include <libavutil/opt.h>
}
#include <algorithm>
#include <atomic>
#include <cassert>
#include <cctype>
#include <cstdlib>
#include <cmath>
#include <cstring>
#include <filesystem>
#include <fstream>
#include <iostream>
#include <iterator>
#include <map>
#include <string>
#include <thread>
#include <vector>
#include <arpa/inet.h>
#include <sys/socket.h>
#include <unistd.h>
#include <fcntl.h>
#include <poll.h>
static void ok(int result) { assert(result >= 0); }
static std::vector<uint8_t> readFile(const std::string &path) {
std::ifstream file(path, std::ios::binary); assert(file.good());
return {std::istreambuf_iterator<char>(file), std::istreambuf_iterator<char>()};
}
static void writeFile(const std::string &path, const std::vector<uint8_t> &bytes) {
std::ofstream file(path, std::ios::binary); file.write(reinterpret_cast<const char *>(bytes.data()), bytes.size()); assert(file.good());
}
// Real source A/V built with the same bundled encoders; no PATH ffmpeg or fixtures.
static void fixture(const std::string &path) {
AVFormatContext *format = nullptr; ok(avformat_alloc_output_context2(&format, nullptr, "mpegts", path.c_str()));
AVCodecContext *video = avcodec_alloc_context3(avcodec_find_encoder_by_name("h264_videotoolbox"));
AVCodecContext *audio = avcodec_alloc_context3(avcodec_find_encoder(AV_CODEC_ID_AAC)); assert(video && audio);
video->width = 96; video->height = 64; video->pix_fmt = AV_PIX_FMT_NV12;
video->time_base = {1, 10}; video->framerate = {10, 1}; video->bit_rate = 300000; video->gop_size = 70; video->max_b_frames = 0;
video->sample_aspect_ratio = {4, 3};
av_opt_set(video->priv_data, "allow_sw", "1", 0); av_opt_set(video->priv_data, "realtime", "1", 0);
audio->sample_rate = 48000; audio->sample_fmt = AV_SAMPLE_FMT_FLTP; audio->time_base = {1, 48000}; audio->bit_rate = 128000;
av_channel_layout_default(&audio->ch_layout, 2);
ok(avcodec_open2(video, video->codec, nullptr)); ok(avcodec_open2(audio, audio->codec, nullptr));
auto *vs = avformat_new_stream(format, nullptr), *as = avformat_new_stream(format, nullptr); assert(vs && as);
vs->time_base = video->time_base; as->time_base = audio->time_base;
ok(avcodec_parameters_from_context(vs->codecpar, video)); ok(avcodec_parameters_from_context(as->codecpar, audio));
AVBSFContext *aspect = nullptr;
ok(av_bsf_alloc(av_bsf_get_by_name("h264_metadata"), &aspect));
ok(avcodec_parameters_copy(aspect->par_in, vs->codecpar));
aspect->time_base_in = video->time_base;
ok(av_opt_set(aspect->priv_data, "sample_aspect_ratio", "4/3", 0));
ok(av_bsf_init(aspect));
ok(avio_open(&format->pb, path.c_str(), AVIO_FLAG_WRITE)); ok(avformat_write_header(format, nullptr));
auto encode = [&](AVCodecContext *codec, AVStream *stream, AVFrame *frame) {
ok(avcodec_send_frame(codec, frame)); auto *packet = av_packet_alloc(); assert(packet);
for (;;) {
int result = avcodec_receive_packet(codec, packet);
if (result == AVERROR(EAGAIN) || result == AVERROR_EOF) break;
ok(result);
if (codec == video) {
ok(av_bsf_send_packet(aspect, packet));
while (av_bsf_receive_packet(aspect, packet) >= 0) {
av_packet_rescale_ts(packet, codec->time_base, stream->time_base); packet->stream_index = stream->index;
ok(av_interleaved_write_frame(format, packet)); av_packet_unref(packet);
}
} else {
av_packet_rescale_ts(packet, codec->time_base, stream->time_base); packet->stream_index = stream->index;
ok(av_interleaved_write_frame(format, packet)); av_packet_unref(packet);
}
}
av_packet_free(&packet);
};
int64_t vi = 0, ai = 0;
while (vi < 900 || ai < 90 * 48000) {
bool isVideo = vi < 900 && (ai >= 90 * 48000 || vi / 10.0 <= ai / 48000.0);
auto *f = av_frame_alloc(); assert(f);
if (isVideo) {
f->width = 96; f->height = 64; f->format = video->pix_fmt; f->pts = vi++;
ok(av_frame_get_buffer(f, 32));
for (int y = 0; y < 64; ++y) std::memset(f->data[0] + y * f->linesize[0], 40 + vi % 160, 96);
for (int y = 0; y < 32; ++y) std::memset(f->data[1] + y * f->linesize[1], 128, 96);
encode(video, vs, f);
} else {
f->sample_rate = 48000; f->format = audio->sample_fmt; f->nb_samples = audio->frame_size;
ok(av_channel_layout_copy(&f->ch_layout, &audio->ch_layout)); f->pts = ai;
ok(av_frame_get_buffer(f, 0));
for (int channel = 0; channel < 2; ++channel) for (int n = 0; n < f->nb_samples; ++n)
reinterpret_cast<float *>(f->data[channel])[n] = float(.1 * std::sin((ai + n) * 440 * 2 * 3.141592653589793 / 48000));
ai += f->nb_samples; encode(audio, as, f);
}
av_frame_free(&f);
}
encode(video, vs, nullptr); encode(audio, as, nullptr); ok(av_write_trailer(format));
av_bsf_free(&aspect);
avio_closep(&format->pb); avformat_free_context(format); avcodec_free_context(&video); avcodec_free_context(&audio);
}
static std::vector<uint8_t> wav() {
std::vector<uint8_t> bytes(44 + 90 * 48000 * 2);
auto put = [&](size_t at, uint32_t value, int length) { for (int i = 0; i < length; ++i) bytes[at + i] = uint8_t(value >> (8 * i)); };
std::memcpy(bytes.data(), "RIFF", 4); put(4, uint32_t(bytes.size() - 8), 4); std::memcpy(bytes.data() + 8, "WAVEfmt ", 8);
put(16, 16, 4); put(20, 1, 2); put(22, 1, 2); put(24, 48000, 4); put(28, 96000, 4); put(32, 2, 2); put(34, 16, 2);
std::memcpy(bytes.data() + 36, "data", 4); put(40, uint32_t(bytes.size() - 44), 4);
for (size_t i = 0; i < 90 * 48000; ++i) put(44 + i * 2, uint16_t(int16_t(4000 * std::sin(i * 440 * 2 * 3.141592653589793 / 48000))), 2);
return bytes;
}
struct HTTP {
int status = 0; std::string headers; std::vector<uint8_t> bytes;
};
static HTTP request(const std::string &url, const std::string &method = "GET", const std::string &headers = "") {
assert(url.starts_with("http://127.0.0.1:")); size_t slash = url.find('/', 7), colon = url.find(':', 7);
sockaddr_in address{}; address.sin_family = AF_INET; address.sin_addr.s_addr = htonl(INADDR_LOOPBACK); address.sin_port = htons(std::stoi(url.substr(colon + 1, slash - colon - 1)));
int socket = ::socket(AF_INET, SOCK_STREAM, 0); assert(socket >= 0);
timeval timeout{60, 0}; setsockopt(socket, SOL_SOCKET, SO_RCVTIMEO, &timeout, sizeof(timeout));
assert(connect(socket, reinterpret_cast<sockaddr *>(&address), sizeof(address)) == 0);
std::string text = method + " " + url.substr(slash) + " HTTP/1.1\r\nHost: 127.0.0.1\r\n" + headers + "Connection: close\r\n\r\n";
assert(send(socket, text.data(), text.size(), 0) == ssize_t(text.size()));
std::vector<uint8_t> bytes; uint8_t buffer[65536];
for (;;) { ssize_t n = recv(socket, buffer, sizeof(buffer), 0); assert(n >= 0); if (!n) break; bytes.insert(bytes.end(), buffer, buffer + n); }
close(socket); std::string raw(bytes.begin(), bytes.end()); size_t end = raw.find("\r\n\r\n"); assert(end != std::string::npos);
return {std::stoi(raw.substr(9, 3)), raw.substr(0, end + 4), {bytes.begin() + end + 4, bytes.end()}};
}
// A tiny real upstream tests resource-specific private headers and byte-range seek.
struct Upstream {
int listener; uint16_t port; std::thread threads[4]; std::atomic<bool> stopped{false};
std::atomic<int> foreignRequests{0}, videoRequests{0}, audioRequests{0};
std::map<std::string, std::vector<uint8_t>> resources;
std::string foreign;
explicit Upstream(std::vector<uint8_t> video, std::vector<uint8_t> audio) {
resources["/source.ts"] = std::move(video); resources["/audio.wav"] = std::move(audio);
resources["/public.ts"] = resources["/source.ts"];
listener = socket(AF_INET, SOCK_STREAM, 0); assert(listener >= 0); int one = 1; setsockopt(listener, SOL_SOCKET, SO_NOSIGPIPE, &one, sizeof(one));
sockaddr_in address{}; address.sin_family = AF_INET; address.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
assert(bind(listener, reinterpret_cast<sockaddr *>(&address), sizeof(address)) == 0); assert(listen(listener, 16) == 0);
assert(fcntl(listener, F_SETFL, O_NONBLOCK) == 0);
socklen_t length = sizeof(address); assert(getsockname(listener, reinterpret_cast<sockaddr *>(&address), &length) == 0); port = ntohs(address.sin_port);
}
std::string url(std::string path) const { return "http://127.0.0.1:" + std::to_string(port) + path; }
void start() { for (auto &thread : threads) thread = std::thread([&] {
while (!stopped) {
pollfd ready{listener, POLLIN, 0};
if (poll(&ready, 1, 100) <= 0 || !(ready.revents & POLLIN)) continue;
int client = accept(listener, nullptr, nullptr); if (client < 0) continue;
assert(fcntl(client, F_SETFL, fcntl(client, F_GETFL) & ~O_NONBLOCK) == 0);
int one = 1; setsockopt(client, SOL_SOCKET, SO_NOSIGPIPE, &one, sizeof(one)); timeval timeout{5, 0}; setsockopt(client, SOL_SOCKET, SO_RCVTIMEO, &timeout, sizeof(timeout)); setsockopt(client, SOL_SOCKET, SO_SNDTIMEO, &timeout, sizeof(timeout));
std::string request; char buffer[2048];
while (request.find("\r\n\r\n") == std::string::npos && request.size() < 16384) { auto n = recv(client, buffer, sizeof(buffer), 0); if (n <= 0) break; request.append(buffer, size_t(n)); }
size_t begin = request.find(' '), end = request.find(' ', begin + 1); std::string path = request.substr(begin + 1, end - begin - 1);
int status = 200; std::string extra; std::vector<uint8_t> data;
if (path == "/redirect") { status = 302; extra = "Location: " + foreign + "/source.ts\r\n"; }
else if (path == "/cross.m3u8" || path == "/index.m3u8" || path == "/public-cross.m3u8") {
std::string playlist = "#EXTM3U\n#EXT-X-TARGETDURATION:90\n#EXTINF:90,\n" +
(path == "/index.m3u8" ? url("") : foreign) +
(path == "/public-cross.m3u8" ? "/public.ts" : "/source.ts") + "\n#EXT-X-ENDLIST\n";
data.assign(playlist.begin(), playlist.end());
} else if (auto resource = resources.find(path); resource != resources.end()) {
bool audio = path == "/audio.wav";
bool publicMedia = path == "/public.ts";
bool authorized = publicMedia || request.find(audio ? "X-Audio: audio-secret" : "X-Video: video-secret") != std::string::npos;
bool wrong = request.find(publicMedia ? "secret" : (audio ? "video-secret" : "audio-secret")) != std::string::npos;
if (!authorized || wrong) { ++foreignRequests; status = 403; }
else { if (audio) ++audioRequests; else ++videoRequests; data = resource->second; }
} else status = 404;
size_t low = 0, high = data.size();
auto range = request.find("Range: bytes=");
if (range != std::string::npos && !data.empty()) {
low = std::stoull(request.substr(range + 13)); auto dash = request.find('-', range);
if (dash != std::string::npos && std::isdigit(request[dash + 1])) high = std::min(high, size_t(std::stoull(request.substr(dash + 1))) + 1);
if (low >= high) { status = 416; low = high = 0; }
else { status = 206; extra += "Content-Range: bytes " + std::to_string(low) + "-" + std::to_string(high - 1) + "/" + std::to_string(data.size()) + "\r\n"; }
}
std::string header = "HTTP/1.1 " + std::to_string(status) + " Test\r\nConnection: close\r\nAccept-Ranges: bytes\r\n" + extra + "Content-Length: " + std::to_string(high - low) + "\r\n\r\n";
send(client, header.data(), header.size(), 0);
if (!request.starts_with("HEAD")) while (high > low) { auto n = send(client, data.data() + low, high - low, 0); if (n <= 0) break; low += size_t(n); }
close(client);
}
}); }
~Upstream() { stopped = true; shutdown(listener, SHUT_RDWR); for (auto &thread : threads) if (thread.joinable()) thread.join(); close(listener); }
};
static void playable(const HTTP &response, const std::string &path, bool video, double minimumTime) {
assert(response.status == 200 && response.bytes.size() > 188 && response.bytes[0] == 0x47); writeFile(path, response.bytes);
AVFormatContext *format = nullptr; ok(avformat_open_input(&format, path.c_str(), nullptr, nullptr)); ok(avformat_find_stream_info(format, nullptr));
int audioIndex = av_find_best_stream(format, AVMEDIA_TYPE_AUDIO, -1, -1, nullptr, 0); assert(audioIndex >= 0);
int videoIndex = av_find_best_stream(format, AVMEDIA_TYPE_VIDEO, -1, -1, nullptr, 0); assert(!video || videoIndex >= 0);
std::map<int, AVCodecContext *> codecs; std::map<int, int> frames;
for (int index : {audioIndex, videoIndex}) if (index >= 0) {
auto *stream = format->streams[index]; auto *codec = avcodec_alloc_context3(avcodec_find_decoder(stream->codecpar->codec_id)); assert(codec);
ok(avcodec_parameters_to_context(codec, stream->codecpar)); ok(avcodec_open2(codec, codec->codec, nullptr)); codecs[index] = codec;
}
auto *packet = av_packet_alloc(); auto *frame = av_frame_alloc(); bool timestamp = false;
while (av_read_frame(format, packet) >= 0) {
auto found = codecs.find(packet->stream_index);
if (found != codecs.end()) {
if (packet->pts != AV_NOPTS_VALUE) { double time = packet->pts * av_q2d(format->streams[packet->stream_index]->time_base); assert(time >= minimumTime - .1); timestamp = true; }
ok(avcodec_send_packet(found->second, packet));
while (avcodec_receive_frame(found->second, frame) >= 0) {
if (packet->stream_index == videoIndex) {
auto sar = av_guess_sample_aspect_ratio(format, format->streams[videoIndex], frame);
if (sar.num <= 0 || sar.den <= 0) sar = {1, 1};
assert(sar.num > 0 && std::abs(double(frame->width) / frame->height * av_q2d(sar) - 2.0) < .01);
}
++frames[packet->stream_index]; av_frame_unref(frame);
}
}
av_packet_unref(packet);
}
assert(timestamp && frames[audioIndex] > 10 && (!video || frames[videoIndex] > 10));
for (auto &[index, codec] : codecs) avcodec_free_context(&codec);
av_packet_free(&packet); av_frame_free(&frame); avformat_close_input(&format);
}
static std::pair<double, double> audioSpan(const HTTP &response, const std::string &path) {
assert(response.status == 200);
writeFile(path, response.bytes);
AVFormatContext *format = nullptr;
ok(avformat_open_input(&format, path.c_str(), nullptr, nullptr));
ok(avformat_find_stream_info(format, nullptr));
int index = av_find_best_stream(format, AVMEDIA_TYPE_AUDIO, -1, -1, nullptr, 0); assert(index >= 0);
auto *codec = avcodec_alloc_context3(avcodec_find_decoder(format->streams[index]->codecpar->codec_id)); assert(codec);
ok(avcodec_parameters_to_context(codec, format->streams[index]->codecpar)); ok(avcodec_open2(codec, codec->codec, nullptr));
auto *packet = av_packet_alloc(); auto *frame = av_frame_alloc();
double first = INFINITY, end = -INFINITY;
while (av_read_frame(format, packet) >= 0) {
if (packet->stream_index == index) {
ok(avcodec_send_packet(codec, packet));
while (avcodec_receive_frame(codec, frame) >= 0) {
assert(frame->best_effort_timestamp != AV_NOPTS_VALUE);
double position = frame->best_effort_timestamp * av_q2d(format->streams[index]->time_base);
first = std::min(first, position);
end = position + double(frame->nb_samples) / frame->sample_rate;
av_frame_unref(frame);
}
}
av_packet_unref(packet);
}
av_packet_free(&packet); av_frame_free(&frame); avcodec_free_context(&codec); avformat_close_input(&format);
assert(std::isfinite(first) && end > first);
return {first, end};
}
static CastMedia *start(CastMediaSource source, std::string &url, double &duration) {
char error[512]{}, bytes[512]{}; auto *media = cast_media_create(&source, error); assert(media);
assert(cast_media_start(media, "127.0.0.1", bytes, &duration, error)); url = bytes; return media;
}
int main(int argc, char **argv) {
bool liveCheck = false;
for (int argument = 1; argument < argc; ++argument) if (std::string(argv[argument]) == "--live") liveCheck = true;
char temp[] = "/tmp/terebi-cast-check.XXXXXX"; assert(mkdtemp(temp)); std::filesystem::path directory(temp);
const auto video = (directory / "source.ts").string(), audio = (directory / "source.wav").string(), segment = (directory / "segment.ts").string();
fixture(video); writeFile(audio, wav());
auto videoBytes = readFile(video), audioBytes = readFile(audio);
Upstream upstream(videoBytes, audioBytes), foreign(videoBytes, audioBytes);
upstream.foreign = foreign.url(""); upstream.start(); foreign.start();
std::string url; double duration = 0;
CastMediaSource local{video.c_str(), "", nullptr, nullptr, true, false, false, 0};
auto *media = start(local, url, duration); assert(duration >= 89 && duration <= 92);
std::string base = url.substr(0, url.rfind('/') + 1);
auto playlist = request(url); assert(playlist.status == 200); std::string text(playlist.bytes.begin(), playlist.bytes.end()); assert(text.find("#EXT-X-ENDLIST") != std::string::npos);
playable(request(base + "segment/10.ts"), segment, true, 60);
playable(request(base + "segment/0.ts"), segment, true, 0);
assert(request(base + "../source.ts").status == 404);
assert(request(base + "%2e%2e/source.ts").status == 404);
assert(request(url.substr(0, url.find('/', 7)) + "/index.m3u8").status == 404);
assert(request(base + "segment/999999.ts").status == 404);
const char vtt[] = "WEBVTT\n\n00:00:01.000 --> 00:00:02.000\nHello\n"; char subtitle[512]{};
assert(cast_media_subtitle(media, 1, vtt, sizeof(vtt) - 1, subtitle));
auto subtitles = request(subtitle); assert(subtitles.status == 200 && subtitles.headers.find("Access-Control-Allow-Origin: *") != std::string::npos);
auto ranged = request(subtitle, "GET", "Range: bytes=0-5\r\n"); assert(ranged.status == 206 && ranged.bytes.size() == 6);
assert(request(subtitle, "HEAD").bytes.empty()); assert(request(subtitle, "OPTIONS").status == 204);
assert(request(subtitle, "GET", "Range: bytes=99999-\r\n").status == 416);
cast_media_cancel(media); cast_media_destroy(media);
// Previously published capabilities expire, including the bound listening socket.
auto authority = url.substr(7, url.find('/', 7) - 7); auto colon = authority.find(':'); sockaddr_in closed{};
closed.sin_family = AF_INET; closed.sin_addr.s_addr = htonl(INADDR_LOOPBACK); closed.sin_port = htons(std::stoi(authority.substr(colon + 1)));
int socket = ::socket(AF_INET, SOCK_STREAM, 0); assert(connect(socket, reinterpret_cast<sockaddr *>(&closed), sizeof(closed)) != 0); close(socket);
std::string hls = upstream.url("/index.m3u8"); CastMediaSource miruro{hls.c_str(), "X-Video: video-secret\r\n", nullptr, nullptr, false, true, false, 90};
media = start(miruro, url, duration); base = url.substr(0, url.rfind('/') + 1);
playable(request(base + "segment/10.ts"), segment, true, 60); cast_media_destroy(media);
std::string networkVideo = upstream.url("/source.ts"), networkAudio = upstream.url("/audio.wav");
CastMediaSource youtube{networkVideo.c_str(), "X-Video: video-secret\r\n", networkAudio.c_str(), "X-Audio: audio-secret\r\n", false, false, false, 90};
media = start(youtube, url, duration); base = url.substr(0, url.rfind('/') + 1);
playable(request(base + "segment/10.ts"), segment, true, 60);
auto vodFirstAudio = audioSpan(request(base + "segment/0.ts"), segment);
auto vodSecondAudio = audioSpan(request(base + "segment/1.ts"), segment);
assert(std::abs(vodSecondAudio.first - vodFirstAudio.second) < 1.0 / 48000);
cast_media_destroy(media);
assert(upstream.videoRequests > 0 && upstream.audioRequests > 0 && upstream.foreignRequests == 0);
CastMediaSource audioOnly{audio.c_str(), "", nullptr, nullptr, true, false, false, 0};
media = start(audioOnly, url, duration); base = url.substr(0, url.rfind('/') + 1); playable(request(base + "segment/10.ts"), segment, false, 60); cast_media_destroy(media);
for (const auto &path : {"/redirect", "/cross.m3u8"}) {
auto resource = upstream.url(path); CastMediaSource unsafe{resource.c_str(), "X-Video: video-secret\r\n", nullptr, nullptr, false, true, false, 90};
char error[512]{}, result[512]{}; media = cast_media_create(&unsafe, error); assert(media);
assert(!cast_media_start(media, "127.0.0.1", result, &duration, error)); assert(std::string(error).find("secret") == std::string::npos); cast_media_destroy(media);
}
assert(foreign.foreignRequests == 0 && foreign.videoRequests == 0);
auto publicManifest = upstream.url("/public-cross.m3u8");
CastMediaSource publicHLS{publicManifest.c_str(), "X-Video: video-secret\r\n", nullptr, nullptr, false, false, false, 90};
media = start(publicHLS, url, duration); base = url.substr(0, url.rfind('/') + 1);
playable(request(base + "segment/0.ts"), segment, true, 0); cast_media_destroy(media);
assert(foreign.videoRequests > 0 && foreign.foreignRequests == 0);
if (liveCheck) {
youtube.live = true; media = start(youtube, url, duration); base = url.substr(0, url.rfind('/') + 1); assert(duration == 0);
auto firstAudio = audioSpan(request(base + "segment/0.ts"), segment);
auto secondAudio = audioSpan(request(base + "segment/1.ts"), segment);
assert(std::abs(secondAudio.first - firstAudio.second) < 1.0 / 48000);
#ifdef CAST_MEDIA_CHECK
assert(cast_media_check_advance(media, 54));
#else
assert(false && "Compile the permanent live check with CAST_MEDIA_CHECK");
#endif
auto response = request(url); assert(response.status == 200); std::string text(response.bytes.begin(), response.bytes.end());
auto at = text.find("#EXT-X-MEDIA-SEQUENCE:"); assert(at != std::string::npos);
int64_t sequence = std::stoll(text.substr(at + 22));
size_t count = 0, cursor = 0;
while ((cursor = text.find("#EXTINF:", cursor)) != std::string::npos) { ++count; cursor += 8; }
assert(count == 8 && text.find("#EXT-X-ENDLIST") == std::string::npos);
int64_t last = sequence + int64_t(count) - 1;
assert(sequence > 0 && request(base + "segment/0.ts").status == 404);
playable(request(base + "segment/" + std::to_string(last) + ".ts"), segment, true, 0); cast_media_destroy(media);
}
std::filesystem::remove_all(directory);
std::cout << "cast-media: playable A/V, VOD random seek, separate audio, HLS origin/header policy, subtitles, ranges, capabilities, cancellation" << (liveCheck ? ", bounded live window" : " (use --live for rolling window)") << " passed\n";
}