tests: Move all test programs to a subdirectory
[libav.git] / libavformat / tests / movenc.c
1 /*
2 * Copyright (c) 2015 Martin Storsjo
3 *
4 * This file is part of Libav.
5 *
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include "config.h"
22
23 #include "libavutil/intreadwrite.h"
24 #include "libavutil/mathematics.h"
25 #include "libavutil/md5.h"
26
27 #include "libavformat/avformat.h"
28
29 #if HAVE_UNISTD_H
30 #include <unistd.h>
31 #endif
32
33 #if !HAVE_GETOPT
34 #include "compat/getopt.c"
35 #endif
36
37 #define HASH_SIZE 16
38
39 static const uint8_t h264_extradata[] = {
40 0x01, 0x4d, 0x40, 0x1e, 0xff, 0xe1, 0x00, 0x02, 0x67, 0x4d, 0x01, 0x00, 0x02, 0x68, 0xef
41 };
42 static const uint8_t aac_extradata[] = {
43 0x12, 0x10
44 };
45
46
47 const char *format = "mp4";
48 AVFormatContext *ctx;
49 uint8_t iobuf[32768];
50 AVDictionary *opts;
51
52 int write_file;
53 const char *cur_name;
54 FILE* out;
55 int out_size;
56 struct AVMD5* md5;
57 uint8_t hash[HASH_SIZE];
58
59 AVStream *video_st, *audio_st;
60 int64_t audio_dts, video_dts;
61
62 int bframes;
63 int64_t duration;
64 int64_t audio_duration;
65 int frames;
66 int gop_size;
67 int64_t next_p_pts;
68 enum AVPictureType last_picture;
69 int skip_write;
70 int skip_write_audio;
71 int clear_duration;
72
73 int num_warnings;
74
75 int check_faults;
76
77
78 static void count_warnings(void *avcl, int level, const char *fmt, va_list vl)
79 {
80 if (level == AV_LOG_WARNING)
81 num_warnings++;
82 }
83
84 static void init_count_warnings(void)
85 {
86 av_log_set_callback(count_warnings);
87 num_warnings = 0;
88 }
89
90 static void reset_count_warnings(void)
91 {
92 av_log_set_callback(av_log_default_callback);
93 }
94
95 static int io_write(void *opaque, uint8_t *buf, int size)
96 {
97 out_size += size;
98 av_md5_update(md5, buf, size);
99 if (out)
100 fwrite(buf, 1, size, out);
101 return size;
102 }
103
104 static void init_out(const char *name)
105 {
106 char buf[100];
107 cur_name = name;
108 snprintf(buf, sizeof(buf), "%s.%s", cur_name, format);
109
110 av_md5_init(md5);
111 if (write_file) {
112 out = fopen(buf, "wb");
113 if (!out)
114 perror(buf);
115 }
116 out_size = 0;
117 }
118
119 static void close_out(void)
120 {
121 int i;
122 av_md5_final(md5, hash);
123 for (i = 0; i < HASH_SIZE; i++)
124 printf("%02x", hash[i]);
125 printf(" %d %s\n", out_size, cur_name);
126 if (out)
127 fclose(out);
128 out = NULL;
129 }
130
131 static void check_func(int value, int line, const char *msg, ...)
132 {
133 if (!value) {
134 va_list ap;
135 va_start(ap, msg);
136 printf("%d: ", line);
137 vprintf(msg, ap);
138 printf("\n");
139 check_faults++;
140 va_end(ap);
141 }
142 }
143 #define check(value, ...) check_func(value, __LINE__, __VA_ARGS__)
144
145 static void init_fps(int bf, int audio_preroll, int fps)
146 {
147 AVStream *st;
148 ctx = avformat_alloc_context();
149 if (!ctx)
150 exit(1);
151 ctx->oformat = av_guess_format(format, NULL, NULL);
152 if (!ctx->oformat)
153 exit(1);
154 ctx->pb = avio_alloc_context(iobuf, sizeof(iobuf), AVIO_FLAG_WRITE, NULL, NULL, io_write, NULL);
155 if (!ctx->pb)
156 exit(1);
157 ctx->flags |= AVFMT_FLAG_BITEXACT;
158
159 st = avformat_new_stream(ctx, NULL);
160 if (!st)
161 exit(1);
162 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
163 st->codecpar->codec_id = AV_CODEC_ID_H264;
164 st->codecpar->width = 640;
165 st->codecpar->height = 480;
166 st->time_base.num = 1;
167 st->time_base.den = 30;
168 st->codecpar->extradata_size = sizeof(h264_extradata);
169 st->codecpar->extradata = av_mallocz(st->codecpar->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
170 if (!st->codecpar->extradata)
171 exit(1);
172 memcpy(st->codecpar->extradata, h264_extradata, sizeof(h264_extradata));
173 video_st = st;
174
175 st = avformat_new_stream(ctx, NULL);
176 if (!st)
177 exit(1);
178 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
179 st->codecpar->codec_id = AV_CODEC_ID_AAC;
180 st->codecpar->sample_rate = 44100;
181 st->codecpar->channels = 2;
182 st->time_base.num = 1;
183 st->time_base.den = 44100;
184 st->codecpar->extradata_size = sizeof(aac_extradata);
185 st->codecpar->extradata = av_mallocz(st->codecpar->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
186 if (!st->codecpar->extradata)
187 exit(1);
188 memcpy(st->codecpar->extradata, aac_extradata, sizeof(aac_extradata));
189 audio_st = st;
190
191 if (avformat_write_header(ctx, &opts) < 0)
192 exit(1);
193 av_dict_free(&opts);
194
195 frames = 0;
196 gop_size = 30;
197 duration = video_st->time_base.den / fps;
198 audio_duration = 1024LL * audio_st->time_base.den / audio_st->codecpar->sample_rate;
199 if (audio_preroll)
200 audio_preroll = 2048LL * audio_st->time_base.den / audio_st->codecpar->sample_rate;
201
202 bframes = bf;
203 video_dts = bframes ? -duration : 0;
204 audio_dts = -audio_preroll;
205 }
206
207 static void init(int bf, int audio_preroll)
208 {
209 init_fps(bf, audio_preroll, 30);
210 }
211
212 static void mux_frames(int n)
213 {
214 int end_frames = frames + n;
215 while (1) {
216 AVPacket pkt;
217 uint8_t pktdata[4];
218 av_init_packet(&pkt);
219
220 if (av_compare_ts(audio_dts, audio_st->time_base, video_dts, video_st->time_base) < 0) {
221 pkt.dts = pkt.pts = audio_dts;
222 pkt.stream_index = 1;
223 pkt.duration = audio_duration;
224 audio_dts += audio_duration;
225 } else {
226 if (frames == end_frames)
227 break;
228 pkt.dts = video_dts;
229 pkt.stream_index = 0;
230 pkt.duration = duration;
231 if ((frames % gop_size) == 0) {
232 pkt.flags |= AV_PKT_FLAG_KEY;
233 last_picture = AV_PICTURE_TYPE_I;
234 pkt.pts = pkt.dts + duration;
235 video_dts = pkt.pts;
236 } else {
237 if (last_picture == AV_PICTURE_TYPE_P) {
238 last_picture = AV_PICTURE_TYPE_B;
239 pkt.pts = pkt.dts;
240 video_dts = next_p_pts;
241 } else {
242 last_picture = AV_PICTURE_TYPE_P;
243 if (((frames + 1) % gop_size) == 0) {
244 pkt.pts = pkt.dts + duration;
245 video_dts = pkt.pts;
246 } else {
247 next_p_pts = pkt.pts = pkt.dts + 2 * duration;
248 video_dts += duration;
249 }
250 }
251 }
252 if (!bframes)
253 pkt.pts = pkt.dts;
254 frames++;
255 }
256
257 if (clear_duration)
258 pkt.duration = 0;
259 AV_WB32(pktdata, pkt.pts);
260 pkt.data = pktdata;
261 pkt.size = 4;
262 if (skip_write)
263 continue;
264 if (skip_write_audio && pkt.stream_index == 1)
265 continue;
266 av_write_frame(ctx, &pkt);
267 }
268 }
269
270 static void mux_gops(int n)
271 {
272 mux_frames(gop_size * n);
273 }
274
275 static void skip_gops(int n)
276 {
277 skip_write = 1;
278 mux_gops(n);
279 skip_write = 0;
280 }
281
282 static void signal_init_ts(void)
283 {
284 AVPacket pkt;
285 av_init_packet(&pkt);
286 pkt.size = 0;
287 pkt.data = NULL;
288
289 pkt.stream_index = 0;
290 pkt.dts = video_dts;
291 pkt.pts = 0;
292 av_write_frame(ctx, &pkt);
293
294 pkt.stream_index = 1;
295 pkt.dts = pkt.pts = audio_dts;
296 av_write_frame(ctx, &pkt);
297 }
298
299 static void finish(void)
300 {
301 av_write_trailer(ctx);
302 av_free(ctx->pb);
303 avformat_free_context(ctx);
304 ctx = NULL;
305 }
306
307 static void help(void)
308 {
309 printf("movenc-test [-w]\n"
310 "-w write output into files\n");
311 }
312
313 int main(int argc, char **argv)
314 {
315 int c;
316 uint8_t header[HASH_SIZE];
317 uint8_t content[HASH_SIZE];
318 int empty_moov_pos;
319 int prev_pos;
320
321 for (;;) {
322 c = getopt(argc, argv, "wh");
323 if (c == -1)
324 break;
325 switch (c) {
326 case 'w':
327 write_file = 1;
328 break;
329 default:
330 case 'h':
331 help();
332 return 0;
333 }
334 }
335
336 av_register_all();
337
338 md5 = av_md5_alloc();
339 if (!md5)
340 return 1;
341
342 // Write a fragmented file with an initial moov that actually contains some
343 // samples. One moov+mdat with 1 second of data and one moof+mdat with 1
344 // second of data.
345 init_out("non-empty-moov");
346 av_dict_set(&opts, "movflags", "frag_keyframe", 0);
347 init(0, 0);
348 mux_gops(2);
349 finish();
350 close_out();
351
352 // Write a similar file, but with B-frames and audio preroll, handled
353 // via an edit list.
354 init_out("non-empty-moov-elst");
355 av_dict_set(&opts, "movflags", "frag_keyframe", 0);
356 av_dict_set(&opts, "use_editlist", "1", 0);
357 init(1, 1);
358 mux_gops(2);
359 finish();
360 close_out();
361
362 // Use B-frames but no audio-preroll, but without an edit list.
363 // Due to avoid_negative_ts == AVFMT_AVOID_NEG_TS_MAKE_ZERO, the dts
364 // of the first audio packet is > 0, but it is set to zero since edit
365 // lists aren't used, increasing the duration of the first packet instead.
366 init_out("non-empty-moov-no-elst");
367 av_dict_set(&opts, "movflags", "frag_keyframe", 0);
368 av_dict_set(&opts, "use_editlist", "0", 0);
369 init(1, 0);
370 mux_gops(2);
371 finish();
372 close_out();
373
374 format = "ismv";
375 // Write an ISMV, with B-frames and audio preroll.
376 init_out("ismv");
377 av_dict_set(&opts, "movflags", "frag_keyframe", 0);
378 init(1, 1);
379 mux_gops(2);
380 finish();
381 close_out();
382 format = "mp4";
383
384 // An initial moov that doesn't contain any samples, followed by two
385 // moof+mdat pairs.
386 init_out("empty-moov");
387 av_dict_set(&opts, "movflags", "frag_keyframe+empty_moov", 0);
388 init(0, 0);
389 mux_gops(2);
390 finish();
391 close_out();
392 memcpy(content, hash, HASH_SIZE);
393
394 // Similar to the previous one, but with input that doesn't start at
395 // pts/dts 0. avoid_negative_ts behaves in the same way as
396 // in non-empty-moov-no-elst above.
397 init_out("empty-moov-no-elst");
398 av_dict_set(&opts, "movflags", "frag_keyframe+empty_moov", 0);
399 init(1, 0);
400 mux_gops(2);
401 finish();
402 close_out();
403
404 // Same as the previous one, but disable avoid_negative_ts (which
405 // would require using an edit list, but with empty_moov, one can't
406 // write a sensible edit list, when the start timestamps aren't known).
407 // This should trigger a warning - we check that the warning is produced.
408 init_count_warnings();
409 init_out("empty-moov-no-elst-no-adjust");
410 av_dict_set(&opts, "movflags", "frag_keyframe+empty_moov", 0);
411 av_dict_set(&opts, "avoid_negative_ts", "0", 0);
412 init(1, 0);
413 mux_gops(2);
414 finish();
415 close_out();
416
417 reset_count_warnings();
418 check(num_warnings > 0, "No warnings printed for unhandled start offset");
419
420 // Verify that delay_moov produces the same as empty_moov for
421 // simple input
422 init_out("delay-moov");
423 av_dict_set(&opts, "movflags", "frag_keyframe+delay_moov", 0);
424 init(0, 0);
425 mux_gops(2);
426 finish();
427 close_out();
428 check(!memcmp(hash, content, HASH_SIZE), "delay_moov differs from empty_moov");
429
430 // Test writing content that requires an edit list using delay_moov
431 init_out("delay-moov-elst");
432 av_dict_set(&opts, "movflags", "frag_keyframe+delay_moov", 0);
433 init(1, 1);
434 mux_gops(2);
435 finish();
436 close_out();
437
438 // Test writing a file with one track lacking packets, with delay_moov.
439 skip_write_audio = 1;
440 init_out("delay-moov-empty-track");
441 av_dict_set(&opts, "movflags", "frag_keyframe+delay_moov", 0);
442 init(0, 0);
443 mux_gops(2);
444 // The automatic flushing shouldn't output anything, since we're still
445 // waiting for data for some tracks
446 check(out_size == 0, "delay_moov flushed prematurely");
447 // When closed (or manually flushed), all the written data should still
448 // be output.
449 finish();
450 close_out();
451 check(out_size > 0, "delay_moov didn't output anything");
452
453 // Check that manually flushing still outputs things as expected. This
454 // produces two fragments, while the one above produces only one.
455 init_out("delay-moov-empty-track-flush");
456 av_dict_set(&opts, "movflags", "frag_custom+delay_moov", 0);
457 init(0, 0);
458 mux_gops(1);
459 av_write_frame(ctx, NULL); // Force writing the moov
460 check(out_size > 0, "No moov written");
461 av_write_frame(ctx, NULL);
462 mux_gops(1);
463 av_write_frame(ctx, NULL);
464 finish();
465 close_out();
466
467 skip_write_audio = 0;
468
469
470
471 // Verify that the header written by delay_moov when manually flushed
472 // is identical to the one by empty_moov.
473 init_out("empty-moov-header");
474 av_dict_set(&opts, "movflags", "frag_keyframe+empty_moov", 0);
475 init(0, 0);
476 close_out();
477 memcpy(header, hash, HASH_SIZE);
478 init_out("empty-moov-content");
479 mux_gops(2);
480 // Written 2 seconds of content, with an automatic flush after 1 second.
481 check(out_size > 0, "No automatic flush?");
482 empty_moov_pos = prev_pos = out_size;
483 // Manually flush the second fragment
484 av_write_frame(ctx, NULL);
485 check(out_size > prev_pos, "No second fragment flushed?");
486 prev_pos = out_size;
487 // Check that an extra flush doesn't output any more data
488 av_write_frame(ctx, NULL);
489 check(out_size == prev_pos, "More data written?");
490 close_out();
491 memcpy(content, hash, HASH_SIZE);
492 // Ignore the trailer written here
493 finish();
494
495 init_out("delay-moov-header");
496 av_dict_set(&opts, "movflags", "frag_custom+delay_moov", 0);
497 init(0, 0);
498 check(out_size == 0, "Output written during init with delay_moov");
499 mux_gops(1); // Write 1 second of content
500 av_write_frame(ctx, NULL); // Force writing the moov
501 close_out();
502 check(!memcmp(hash, header, HASH_SIZE), "delay_moov header differs from empty_moov");
503 init_out("delay-moov-content");
504 av_write_frame(ctx, NULL); // Flush the first fragment
505 check(out_size == empty_moov_pos, "Manually flushed content differs from automatically flushed, %d vs %d", out_size, empty_moov_pos);
506 mux_gops(1); // Write the rest of the content
507 av_write_frame(ctx, NULL); // Flush the second fragment
508 close_out();
509 check(!memcmp(hash, content, HASH_SIZE), "delay_moov content differs from empty_moov");
510 finish();
511
512
513 // Verify that we can produce an identical second fragment without
514 // writing the first one. First write the reference fragments that
515 // we want to reproduce.
516 av_dict_set(&opts, "movflags", "frag_custom+empty_moov+dash", 0);
517 init(0, 0);
518 mux_gops(1);
519 av_write_frame(ctx, NULL); // Output the first fragment
520 init_out("empty-moov-second-frag");
521 mux_gops(1);
522 av_write_frame(ctx, NULL); // Output the second fragment
523 close_out();
524 memcpy(content, hash, HASH_SIZE);
525 finish();
526
527 // Produce the same second fragment without actually writing the first
528 // one before.
529 av_dict_set(&opts, "movflags", "frag_custom+empty_moov+dash+frag_discont", 0);
530 av_dict_set(&opts, "fragment_index", "2", 0);
531 av_dict_set(&opts, "avoid_negative_ts", "0", 0);
532 av_dict_set(&opts, "use_editlist", "0", 0);
533 init(0, 0);
534 skip_gops(1);
535 init_out("empty-moov-second-frag-discont");
536 mux_gops(1);
537 av_write_frame(ctx, NULL); // Output the second fragment
538 close_out();
539 check(!memcmp(hash, content, HASH_SIZE), "discontinuously written fragment differs");
540 finish();
541
542 // Produce the same thing by using delay_moov, which requires a slightly
543 // different call sequence.
544 av_dict_set(&opts, "movflags", "frag_custom+delay_moov+dash+frag_discont", 0);
545 av_dict_set(&opts, "fragment_index", "2", 0);
546 init(0, 0);
547 skip_gops(1);
548 mux_gops(1);
549 av_write_frame(ctx, NULL); // Output the moov
550 init_out("delay-moov-second-frag-discont");
551 av_write_frame(ctx, NULL); // Output the second fragment
552 close_out();
553 check(!memcmp(hash, content, HASH_SIZE), "discontinuously written fragment differs");
554 finish();
555
556
557 // Test discontinuously written fragments with B-frames (where the
558 // assumption of starting at pts=0 works) but not with audio preroll
559 // (which can't be guessed).
560 av_dict_set(&opts, "movflags", "frag_custom+delay_moov+dash", 0);
561 init(1, 0);
562 mux_gops(1);
563 init_out("delay-moov-elst-init");
564 av_write_frame(ctx, NULL); // Output the moov
565 close_out();
566 memcpy(header, hash, HASH_SIZE);
567 av_write_frame(ctx, NULL); // Output the first fragment
568 init_out("delay-moov-elst-second-frag");
569 mux_gops(1);
570 av_write_frame(ctx, NULL); // Output the second fragment
571 close_out();
572 memcpy(content, hash, HASH_SIZE);
573 finish();
574
575 av_dict_set(&opts, "movflags", "frag_custom+delay_moov+dash+frag_discont", 0);
576 av_dict_set(&opts, "fragment_index", "2", 0);
577 init(1, 0);
578 skip_gops(1);
579 mux_gops(1); // Write the second fragment
580 init_out("delay-moov-elst-init-discont");
581 av_write_frame(ctx, NULL); // Output the moov
582 close_out();
583 check(!memcmp(hash, header, HASH_SIZE), "discontinuously written header differs");
584 init_out("delay-moov-elst-second-frag-discont");
585 av_write_frame(ctx, NULL); // Output the second fragment
586 close_out();
587 check(!memcmp(hash, content, HASH_SIZE), "discontinuously written fragment differs");
588 finish();
589
590
591 // Test discontinuously written fragments with B-frames and audio preroll,
592 // properly signaled.
593 av_dict_set(&opts, "movflags", "frag_custom+delay_moov+dash", 0);
594 init(1, 1);
595 mux_gops(1);
596 init_out("delay-moov-elst-signal-init");
597 av_write_frame(ctx, NULL); // Output the moov
598 close_out();
599 memcpy(header, hash, HASH_SIZE);
600 av_write_frame(ctx, NULL); // Output the first fragment
601 init_out("delay-moov-elst-signal-second-frag");
602 mux_gops(1);
603 av_write_frame(ctx, NULL); // Output the second fragment
604 close_out();
605 memcpy(content, hash, HASH_SIZE);
606 finish();
607
608 av_dict_set(&opts, "movflags", "frag_custom+delay_moov+dash+frag_discont", 0);
609 av_dict_set(&opts, "fragment_index", "2", 0);
610 init(1, 1);
611 signal_init_ts();
612 skip_gops(1);
613 mux_gops(1); // Write the second fragment
614 init_out("delay-moov-elst-signal-init-discont");
615 av_write_frame(ctx, NULL); // Output the moov
616 close_out();
617 check(!memcmp(hash, header, HASH_SIZE), "discontinuously written header differs");
618 init_out("delay-moov-elst-signal-second-frag-discont");
619 av_write_frame(ctx, NULL); // Output the second fragment
620 close_out();
621 check(!memcmp(hash, content, HASH_SIZE), "discontinuously written fragment differs");
622 finish();
623
624
625 // Test VFR content, with sidx atoms (which declare the pts duration
626 // of a fragment, forcing overriding the start pts of the next one).
627 // Here, the fragment duration in pts is significantly different from
628 // the duration in dts. The video stream starts at dts=-10,pts=0, and
629 // the second fragment starts at dts=155,pts=156. The trun duration sum
630 // of the first fragment is 165, which also is written as
631 // baseMediaDecodeTime in the tfdt in the second fragment. The sidx for
632 // the first fragment says earliest_presentation_time = 0 and
633 // subsegment_duration = 156, which also matches the sidx in the second
634 // fragment. For the audio stream, the pts and dts durations also don't
635 // match - the input stream starts at pts=-2048, but that part is excluded
636 // by the edit list.
637 init_out("vfr");
638 av_dict_set(&opts, "movflags", "frag_keyframe+delay_moov+dash", 0);
639 init_fps(1, 1, 3);
640 mux_frames(gop_size/2);
641 duration /= 10;
642 mux_frames(gop_size/2);
643 mux_gops(1);
644 finish();
645 close_out();
646
647 // Test VFR content, with cleared duration fields. In these cases,
648 // the muxer must guess the duration of the last packet of each
649 // fragment. As long as the framerate doesn't vary (too much) at the
650 // fragment edge, it works just fine. Additionally, when automatically
651 // cutting fragments, the muxer already know the timestamps of the next
652 // packet for one stream (in most cases the video stream), avoiding
653 // having to use guesses for that one.
654 init_count_warnings();
655 clear_duration = 1;
656 init_out("vfr-noduration");
657 av_dict_set(&opts, "movflags", "frag_keyframe+delay_moov+dash", 0);
658 init_fps(1, 1, 3);
659 mux_frames(gop_size/2);
660 duration /= 10;
661 mux_frames(gop_size/2);
662 mux_gops(1);
663 finish();
664 close_out();
665 clear_duration = 0;
666 reset_count_warnings();
667 check(num_warnings > 0, "No warnings printed for filled in durations");
668
669 av_free(md5);
670
671 return check_faults > 0 ? 1 : 0;
672 }