1 | /********** |
2 | This library is free software; you can redistribute it and/or modify it under |
3 | the terms of the GNU Lesser General Public License as published by the |
4 | Free Software Foundation; either version 3 of the License, or (at your |
5 | option) any later version. (See <http://www.gnu.org/copyleft/lesser.html>.) |
6 | |
7 | This library is distributed in the hope that it will be useful, but WITHOUT |
8 | ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS |
9 | FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for |
10 | more details. |
11 | |
12 | You should have received a copy of the GNU Lesser General Public License |
13 | along with this library; if not, write to the Free Software Foundation, Inc., |
14 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
15 | **********/ |
16 | // "liveMedia" |
17 | // Copyright (c) 1996-2020 Live Networks, Inc. All rights reserved. |
18 | // A sink that generates a QuickTime file from a composite media session |
19 | // Implementation |
20 | |
21 | #include "QuickTimeFileSink.hh" |
22 | #include "QuickTimeGenericRTPSource.hh" |
23 | #include "GroupsockHelper.hh" |
24 | #include "InputFile.hh" |
25 | #include "OutputFile.hh" |
26 | #include "H263plusVideoRTPSource.hh" // for the special header |
27 | #include "MPEG4GenericRTPSource.hh" //for "samplingFrequencyFromAudioSpecificConfig()" |
28 | #include "MPEG4LATMAudioRTPSource.hh" // for "parseGeneralConfigStr()" |
29 | #include "Base64.hh" |
30 | |
31 | #include <ctype.h> |
32 | |
33 | #define fourChar(x,y,z,w) ( ((x)<<24)|((y)<<16)|((z)<<8)|(w) ) |
34 | |
35 | #define H264_IDR_FRAME 0x65 //bit 8 == 0, bits 7-6 (ref) == 3, bits 5-0 (type) == 5 |
36 | |
37 | ////////// SubsessionIOState, ChunkDescriptor /////////// |
38 | // A structure used to represent the I/O state of each input 'subsession': |
39 | |
40 | class ChunkDescriptor { |
41 | public: |
42 | ChunkDescriptor(int64_t offsetInFile, unsigned size, |
43 | unsigned frameSize, unsigned frameDuration, |
44 | struct timeval presentationTime); |
45 | |
46 | ChunkDescriptor* extendChunk(int64_t newOffsetInFile, unsigned newSize, |
47 | unsigned newFrameSize, |
48 | unsigned newFrameDuration, |
49 | struct timeval newPresentationTime); |
50 | // this may end up allocating a new chunk instead |
51 | public: |
52 | ChunkDescriptor* fNextChunk; |
53 | int64_t fOffsetInFile; |
54 | unsigned fNumFrames; |
55 | unsigned fFrameSize; |
56 | unsigned fFrameDuration; |
57 | struct timeval fPresentationTime; // of the start of the data |
58 | }; |
59 | |
60 | class SubsessionBuffer { |
61 | public: |
62 | SubsessionBuffer(unsigned bufferSize) |
63 | : fBufferSize(bufferSize) { |
64 | reset(); |
65 | fData = new unsigned char[bufferSize]; |
66 | } |
67 | virtual ~SubsessionBuffer() { delete[] fData; } |
68 | void reset() { fBytesInUse = 0; } |
69 | void addBytes(unsigned numBytes) { fBytesInUse += numBytes; } |
70 | |
71 | unsigned char* dataStart() { return &fData[0]; } |
72 | unsigned char* dataEnd() { return &fData[fBytesInUse]; } |
73 | unsigned bytesInUse() const { return fBytesInUse; } |
74 | unsigned bytesAvailable() const { return fBufferSize - fBytesInUse; } |
75 | |
76 | void setPresentationTime(struct timeval const& presentationTime) { |
77 | fPresentationTime = presentationTime; |
78 | } |
79 | struct timeval const& presentationTime() const {return fPresentationTime;} |
80 | |
81 | private: |
82 | unsigned fBufferSize; |
83 | struct timeval fPresentationTime; |
84 | unsigned char* fData; |
85 | unsigned fBytesInUse; |
86 | }; |
87 | |
88 | class SyncFrame { |
89 | public: |
90 | SyncFrame(unsigned ); |
91 | |
92 | public: |
93 | class SyncFrame *nextSyncFrame; |
94 | unsigned ; |
95 | }; |
96 | |
97 | // A 64-bit counter, used below: |
98 | class Count64 { |
99 | public: |
100 | Count64() |
101 | : hi(0), lo(0) { |
102 | } |
103 | |
104 | void operator+=(unsigned arg); |
105 | |
106 | u_int32_t hi, lo; |
107 | }; |
108 | |
109 | class SubsessionIOState { |
110 | public: |
111 | SubsessionIOState(QuickTimeFileSink& sink, MediaSubsession& subsession); |
112 | virtual ~SubsessionIOState(); |
113 | |
114 | Boolean setQTstate(); |
115 | void setFinalQTstate(); |
116 | |
117 | void afterGettingFrame(unsigned packetDataSize, |
118 | struct timeval presentationTime); |
119 | void onSourceClosure(); |
120 | |
121 | Boolean syncOK(struct timeval presentationTime); |
122 | // returns true iff data is usable despite a sync check |
123 | |
124 | static void setHintTrack(SubsessionIOState* hintedTrack, |
125 | SubsessionIOState* hintTrack); |
126 | Boolean isHintTrack() const { return fTrackHintedByUs != NULL; } |
127 | Boolean hasHintTrack() const { return fHintTrackForUs != NULL; } |
128 | |
129 | UsageEnvironment& envir() const { return fOurSink.envir(); } |
130 | |
131 | public: |
132 | static unsigned fCurrentTrackNumber; |
133 | unsigned fTrackID; |
134 | SubsessionIOState* fHintTrackForUs; SubsessionIOState* fTrackHintedByUs; |
135 | |
136 | SubsessionBuffer *fBuffer, *fPrevBuffer; |
137 | QuickTimeFileSink& fOurSink; |
138 | MediaSubsession& fOurSubsession; |
139 | |
140 | unsigned short fLastPacketRTPSeqNum; |
141 | Boolean fOurSourceIsActive; |
142 | |
143 | Boolean fHaveBeenSynced; // used in synchronizing with other streams |
144 | struct timeval fSyncTime; |
145 | |
146 | Boolean fQTEnableTrack; |
147 | unsigned fQTcomponentSubtype; |
148 | char const* fQTcomponentName; |
149 | typedef unsigned (QuickTimeFileSink::*atomCreationFunc)(); |
150 | atomCreationFunc fQTMediaInformationAtomCreator; |
151 | atomCreationFunc fQTMediaDataAtomCreator; |
152 | char const* fQTAudioDataType; |
153 | unsigned short fQTSoundSampleVersion; |
154 | unsigned fQTTimeScale; |
155 | unsigned fQTTimeUnitsPerSample; |
156 | unsigned fQTBytesPerFrame; |
157 | unsigned fQTSamplesPerFrame; |
158 | // These next fields are derived from the ones above, |
159 | // plus the information from each chunk: |
160 | unsigned fQTTotNumSamples; |
161 | unsigned fQTDurationM; // in media time units |
162 | unsigned fQTDurationT; // in track time units |
163 | int64_t fTKHD_durationPosn; |
164 | // position of the duration in the output 'tkhd' atom |
165 | unsigned fQTInitialOffsetDuration; |
166 | // if there's a pause at the beginning |
167 | |
168 | ChunkDescriptor *fHeadChunk, *fTailChunk; |
169 | unsigned fNumChunks; |
170 | SyncFrame *fHeadSyncFrame, *fTailSyncFrame; |
171 | |
172 | // Counters to be used in the hint track's 'udta'/'hinf' atom; |
173 | struct hinf { |
174 | Count64 trpy; |
175 | Count64 nump; |
176 | Count64 tpyl; |
177 | // Is 'maxr' needed? Computing this would be a PITA. ##### |
178 | Count64 dmed; |
179 | Count64 dimm; |
180 | // 'drep' is always 0 |
181 | // 'tmin' and 'tmax' are always 0 |
182 | unsigned pmax; |
183 | unsigned dmax; |
184 | } fHINF; |
185 | |
186 | private: |
187 | void useFrame(SubsessionBuffer& buffer); |
188 | void useFrameForHinting(unsigned frameSize, |
189 | struct timeval presentationTime, |
190 | unsigned startSampleNumber); |
191 | |
192 | // used by the above two routines: |
193 | unsigned useFrame1(unsigned sourceDataSize, |
194 | struct timeval presentationTime, |
195 | unsigned frameDuration, int64_t destFileOffset); |
196 | // returns the number of samples in this data |
197 | |
198 | private: |
199 | // A structure used for temporarily storing frame state: |
200 | struct { |
201 | unsigned frameSize; |
202 | struct timeval presentationTime; |
203 | int64_t destFileOffset; // used for non-hint tracks only |
204 | |
205 | // The remaining fields are used for hint tracks only: |
206 | unsigned startSampleNumber; |
207 | unsigned short seqNum; |
208 | unsigned ; |
209 | unsigned char ; // used when our RTP source has special headers |
210 | unsigned ; // ditto |
211 | unsigned char [SPECIAL_HEADER_BUFFER_SIZE]; // ditto |
212 | unsigned packetSizes[256]; |
213 | } fPrevFrameState; |
214 | }; |
215 | |
216 | |
217 | ////////// QuickTimeFileSink implementation ////////// |
218 | |
219 | QuickTimeFileSink::QuickTimeFileSink(UsageEnvironment& env, |
220 | MediaSession& inputSession, |
221 | char const* outputFileName, |
222 | unsigned bufferSize, |
223 | unsigned short movieWidth, |
224 | unsigned short movieHeight, |
225 | unsigned movieFPS, |
226 | Boolean packetLossCompensate, |
227 | Boolean syncStreams, |
228 | Boolean generateHintTracks, |
229 | Boolean generateMP4Format) |
230 | : Medium(env), fInputSession(inputSession), |
231 | fBufferSize(bufferSize), fPacketLossCompensate(packetLossCompensate), |
232 | fSyncStreams(syncStreams), fGenerateMP4Format(generateMP4Format), |
233 | fAreCurrentlyBeingPlayed(False), |
234 | fLargestRTPtimestampFrequency(0), |
235 | fNumSubsessions(0), fNumSyncedSubsessions(0), |
236 | fHaveCompletedOutputFile(False), |
237 | fMovieWidth(movieWidth), fMovieHeight(movieHeight), |
238 | fMovieFPS(movieFPS), fMaxTrackDurationM(0) { |
239 | fOutFid = OpenOutputFile(env, outputFileName); |
240 | if (fOutFid == NULL) return; |
241 | |
242 | fNewestSyncTime.tv_sec = fNewestSyncTime.tv_usec = 0; |
243 | fFirstDataTime.tv_sec = fFirstDataTime.tv_usec = (unsigned)(~0); |
244 | |
245 | // Set up I/O state for each input subsession: |
246 | MediaSubsessionIterator iter(fInputSession); |
247 | MediaSubsession* subsession; |
248 | while ((subsession = iter.next()) != NULL) { |
249 | // Ignore subsessions without a data source: |
250 | FramedSource* subsessionSource = subsession->readSource(); |
251 | if (subsessionSource == NULL) continue; |
252 | |
253 | // If "subsession's" SDP description specified screen dimension |
254 | // or frame rate parameters, then use these. (Note that this must |
255 | // be done before the call to "setQTState()" below.) |
256 | if (subsession->videoWidth() != 0) { |
257 | fMovieWidth = subsession->videoWidth(); |
258 | } |
259 | if (subsession->videoHeight() != 0) { |
260 | fMovieHeight = subsession->videoHeight(); |
261 | } |
262 | if (subsession->videoFPS() != 0) { |
263 | fMovieFPS = subsession->videoFPS(); |
264 | } |
265 | |
266 | SubsessionIOState* ioState |
267 | = new SubsessionIOState(*this, *subsession); |
268 | if (ioState == NULL || !ioState->setQTstate()) { |
269 | // We're not able to output a QuickTime track for this subsession |
270 | delete ioState; ioState = NULL; |
271 | continue; |
272 | } |
273 | subsession->miscPtr = (void*)ioState; |
274 | |
275 | if (generateHintTracks) { |
276 | // Also create a hint track for this track: |
277 | SubsessionIOState* hintTrack |
278 | = new SubsessionIOState(*this, *subsession); |
279 | SubsessionIOState::setHintTrack(ioState, hintTrack); |
280 | if (!hintTrack->setQTstate()) { |
281 | delete hintTrack; |
282 | SubsessionIOState::setHintTrack(ioState, NULL); |
283 | } |
284 | } |
285 | |
286 | // Also set a 'BYE' handler for this subsession's RTCP instance: |
287 | if (subsession->rtcpInstance() != NULL) { |
288 | subsession->rtcpInstance()->setByeHandler(onRTCPBye, ioState); |
289 | } |
290 | |
291 | unsigned rtpTimestampFrequency = subsession->rtpTimestampFrequency(); |
292 | if (rtpTimestampFrequency > fLargestRTPtimestampFrequency) { |
293 | fLargestRTPtimestampFrequency = rtpTimestampFrequency; |
294 | } |
295 | |
296 | ++fNumSubsessions; |
297 | } |
298 | |
299 | // Use the current time as the file's creation and modification |
300 | // time. Use Apple's time format: seconds (UTC) since January 1, 1904 |
301 | |
302 | gettimeofday(&fStartTime, NULL); |
303 | fAppleCreationTime = fStartTime.tv_sec - 0x83da4f80; |
304 | |
305 | // Begin by writing a "mdat" atom at the start of the file. |
306 | // (Later, when we've finished copying data to the file, we'll come |
307 | // back and fill in its size.) |
308 | fMDATposition = TellFile64(fOutFid); |
309 | addAtomHeader64("mdat" ); |
310 | // add 64Bit offset |
311 | fMDATposition += 8; |
312 | } |
313 | |
314 | QuickTimeFileSink::~QuickTimeFileSink() { |
315 | completeOutputFile(); |
316 | |
317 | // Then, stop streaming and delete each active "SubsessionIOState": |
318 | MediaSubsessionIterator iter(fInputSession); |
319 | MediaSubsession* subsession; |
320 | while ((subsession = iter.next()) != NULL) { |
321 | if (subsession->readSource() != NULL) subsession->readSource()->stopGettingFrames(); |
322 | |
323 | SubsessionIOState* ioState |
324 | = (SubsessionIOState*)(subsession->miscPtr); |
325 | if (ioState == NULL) continue; |
326 | |
327 | delete ioState->fHintTrackForUs; // if any |
328 | delete ioState; |
329 | } |
330 | |
331 | // Finally, close our output file: |
332 | CloseOutputFile(fOutFid); |
333 | } |
334 | |
335 | QuickTimeFileSink* |
336 | QuickTimeFileSink::createNew(UsageEnvironment& env, |
337 | MediaSession& inputSession, |
338 | char const* outputFileName, |
339 | unsigned bufferSize, |
340 | unsigned short movieWidth, |
341 | unsigned short movieHeight, |
342 | unsigned movieFPS, |
343 | Boolean packetLossCompensate, |
344 | Boolean syncStreams, |
345 | Boolean generateHintTracks, |
346 | Boolean generateMP4Format) { |
347 | QuickTimeFileSink* newSink = |
348 | new QuickTimeFileSink(env, inputSession, outputFileName, bufferSize, movieWidth, movieHeight, movieFPS, |
349 | packetLossCompensate, syncStreams, generateHintTracks, generateMP4Format); |
350 | if (newSink == NULL || newSink->fOutFid == NULL) { |
351 | Medium::close(newSink); |
352 | return NULL; |
353 | } |
354 | |
355 | return newSink; |
356 | } |
357 | |
358 | void QuickTimeFileSink |
359 | ::noteRecordedFrame(MediaSubsession& /*inputSubsession*/, |
360 | unsigned /*packetDataSize*/, struct timeval const& /*presentationTime*/) { |
361 | // Default implementation: Do nothing |
362 | } |
363 | |
364 | Boolean QuickTimeFileSink::startPlaying(afterPlayingFunc* afterFunc, |
365 | void* afterClientData) { |
366 | // Make sure we're not already being played: |
367 | if (fAreCurrentlyBeingPlayed) { |
368 | envir().setResultMsg("This sink has already been played" ); |
369 | return False; |
370 | } |
371 | |
372 | fAreCurrentlyBeingPlayed = True; |
373 | fAfterFunc = afterFunc; |
374 | fAfterClientData = afterClientData; |
375 | |
376 | return continuePlaying(); |
377 | } |
378 | |
379 | Boolean QuickTimeFileSink::continuePlaying() { |
380 | // Run through each of our input session's 'subsessions', |
381 | // asking for a frame from each one: |
382 | Boolean haveActiveSubsessions = False; |
383 | MediaSubsessionIterator iter(fInputSession); |
384 | MediaSubsession* subsession; |
385 | while ((subsession = iter.next()) != NULL) { |
386 | FramedSource* subsessionSource = subsession->readSource(); |
387 | if (subsessionSource == NULL) continue; |
388 | |
389 | if (subsessionSource->isCurrentlyAwaitingData()) continue; |
390 | |
391 | SubsessionIOState* ioState |
392 | = (SubsessionIOState*)(subsession->miscPtr); |
393 | if (ioState == NULL) continue; |
394 | |
395 | haveActiveSubsessions = True; |
396 | unsigned char* toPtr = ioState->fBuffer->dataEnd(); |
397 | unsigned toSize = ioState->fBuffer->bytesAvailable(); |
398 | subsessionSource->getNextFrame(toPtr, toSize, |
399 | afterGettingFrame, ioState, |
400 | onSourceClosure, ioState); |
401 | } |
402 | if (!haveActiveSubsessions) { |
403 | envir().setResultMsg("No subsessions are currently active" ); |
404 | return False; |
405 | } |
406 | |
407 | return True; |
408 | } |
409 | |
410 | void QuickTimeFileSink |
411 | ::afterGettingFrame(void* clientData, unsigned packetDataSize, |
412 | unsigned numTruncatedBytes, |
413 | struct timeval presentationTime, |
414 | unsigned /*durationInMicroseconds*/) { |
415 | SubsessionIOState* ioState = (SubsessionIOState*)clientData; |
416 | if (!ioState->syncOK(presentationTime)) { |
417 | // Ignore this data: |
418 | ioState->fOurSink.continuePlaying(); |
419 | return; |
420 | } |
421 | if (numTruncatedBytes > 0) { |
422 | ioState->envir() << "QuickTimeFileSink::afterGettingFrame(): The input frame data was too large for our buffer. " |
423 | << numTruncatedBytes |
424 | << " bytes of trailing data was dropped! Correct this by increasing the \"bufferSize\" parameter in the \"createNew()\" call.\n" ; |
425 | } |
426 | ioState->afterGettingFrame(packetDataSize, presentationTime); |
427 | } |
428 | |
429 | void QuickTimeFileSink::onSourceClosure(void* clientData) { |
430 | SubsessionIOState* ioState = (SubsessionIOState*)clientData; |
431 | ioState->onSourceClosure(); |
432 | } |
433 | |
434 | void QuickTimeFileSink::onSourceClosure1() { |
435 | // Check whether *all* of the subsession sources have closed. |
436 | // If not, do nothing for now: |
437 | MediaSubsessionIterator iter(fInputSession); |
438 | MediaSubsession* subsession; |
439 | while ((subsession = iter.next()) != NULL) { |
440 | SubsessionIOState* ioState |
441 | = (SubsessionIOState*)(subsession->miscPtr); |
442 | if (ioState == NULL) continue; |
443 | |
444 | if (ioState->fOurSourceIsActive) return; // this source hasn't closed |
445 | } |
446 | |
447 | completeOutputFile(); |
448 | |
449 | // Call our specified 'after' function: |
450 | if (fAfterFunc != NULL) { |
451 | (*fAfterFunc)(fAfterClientData); |
452 | } |
453 | } |
454 | |
455 | void QuickTimeFileSink::onRTCPBye(void* clientData) { |
456 | SubsessionIOState* ioState = (SubsessionIOState*)clientData; |
457 | |
458 | struct timeval timeNow; |
459 | gettimeofday(&timeNow, NULL); |
460 | unsigned secsDiff |
461 | = timeNow.tv_sec - ioState->fOurSink.fStartTime.tv_sec; |
462 | |
463 | MediaSubsession& subsession = ioState->fOurSubsession; |
464 | ioState->envir() << "Received RTCP \"BYE\" on \"" |
465 | << subsession.mediumName() |
466 | << "/" << subsession.codecName() |
467 | << "\" subsession (after " |
468 | << secsDiff << " seconds)\n" ; |
469 | |
470 | // Handle the reception of a RTCP "BYE" as if the source had closed: |
471 | ioState->onSourceClosure(); |
472 | } |
473 | |
474 | static Boolean timevalGE(struct timeval const& tv1, |
475 | struct timeval const& tv2) { |
476 | return (unsigned)tv1.tv_sec > (unsigned)tv2.tv_sec |
477 | || (tv1.tv_sec == tv2.tv_sec |
478 | && (unsigned)tv1.tv_usec >= (unsigned)tv2.tv_usec); |
479 | } |
480 | |
481 | void QuickTimeFileSink::completeOutputFile() { |
482 | if (fHaveCompletedOutputFile || fOutFid == NULL) return; |
483 | |
484 | // Begin by filling in the initial "mdat" atom with the current |
485 | // file size: |
486 | int64_t curFileSize = TellFile64(fOutFid); |
487 | setWord64(fMDATposition, (u_int64_t)curFileSize); |
488 | |
489 | // Then, note the time of the first received data: |
490 | MediaSubsessionIterator iter(fInputSession); |
491 | MediaSubsession* subsession; |
492 | while ((subsession = iter.next()) != NULL) { |
493 | SubsessionIOState* ioState |
494 | = (SubsessionIOState*)(subsession->miscPtr); |
495 | if (ioState == NULL) continue; |
496 | |
497 | ChunkDescriptor* const headChunk = ioState->fHeadChunk; |
498 | if (headChunk != NULL |
499 | && timevalGE(fFirstDataTime, headChunk->fPresentationTime)) { |
500 | fFirstDataTime = headChunk->fPresentationTime; |
501 | } |
502 | } |
503 | |
504 | // Then, update the QuickTime-specific state for each active track: |
505 | iter.reset(); |
506 | while ((subsession = iter.next()) != NULL) { |
507 | SubsessionIOState* ioState |
508 | = (SubsessionIOState*)(subsession->miscPtr); |
509 | if (ioState == NULL) continue; |
510 | |
511 | ioState->setFinalQTstate(); |
512 | // Do the same for a hint track (if any): |
513 | if (ioState->hasHintTrack()) { |
514 | ioState->fHintTrackForUs->setFinalQTstate(); |
515 | } |
516 | } |
517 | |
518 | if (fGenerateMP4Format) { |
519 | // Begin with a "ftyp" atom: |
520 | addAtom_ftyp(); |
521 | } |
522 | |
523 | // Then, add a "moov" atom for the file metadata: |
524 | addAtom_moov(); |
525 | |
526 | // We're done: |
527 | fHaveCompletedOutputFile = True; |
528 | } |
529 | |
530 | |
531 | ////////// SubsessionIOState, ChunkDescriptor implementation /////////// |
532 | |
533 | unsigned SubsessionIOState::fCurrentTrackNumber = 0; |
534 | |
535 | SubsessionIOState::SubsessionIOState(QuickTimeFileSink& sink, |
536 | MediaSubsession& subsession) |
537 | : fHintTrackForUs(NULL), fTrackHintedByUs(NULL), |
538 | fOurSink(sink), fOurSubsession(subsession), |
539 | fLastPacketRTPSeqNum(0), fHaveBeenSynced(False), fQTTotNumSamples(0), |
540 | fHeadChunk(NULL), fTailChunk(NULL), fNumChunks(0), |
541 | fHeadSyncFrame(NULL), fTailSyncFrame(NULL) { |
542 | fTrackID = ++fCurrentTrackNumber; |
543 | |
544 | fBuffer = new SubsessionBuffer(fOurSink.fBufferSize); |
545 | fPrevBuffer = sink.fPacketLossCompensate |
546 | ? new SubsessionBuffer(fOurSink.fBufferSize) : NULL; |
547 | |
548 | FramedSource* subsessionSource = subsession.readSource(); |
549 | fOurSourceIsActive = subsessionSource != NULL; |
550 | |
551 | fPrevFrameState.presentationTime.tv_sec = 0; |
552 | fPrevFrameState.presentationTime.tv_usec = 0; |
553 | fPrevFrameState.seqNum = 0; |
554 | } |
555 | |
556 | SubsessionIOState::~SubsessionIOState() { |
557 | delete fBuffer; delete fPrevBuffer; |
558 | |
559 | // Delete the list of chunk descriptors: |
560 | ChunkDescriptor* chunk = fHeadChunk; |
561 | while (chunk != NULL) { |
562 | ChunkDescriptor* next = chunk->fNextChunk; |
563 | delete chunk; |
564 | chunk = next; |
565 | } |
566 | |
567 | // Delete the list of sync frames: |
568 | SyncFrame* syncFrame = fHeadSyncFrame; |
569 | while (syncFrame != NULL) { |
570 | SyncFrame* next = syncFrame->nextSyncFrame; |
571 | delete syncFrame; |
572 | syncFrame = next; |
573 | } |
574 | } |
575 | |
576 | Boolean SubsessionIOState::setQTstate() { |
577 | char const* noCodecWarning1 = "Warning: We don't implement a QuickTime " ; |
578 | char const* noCodecWarning2 = " Media Data Type for the \"" ; |
579 | char const* noCodecWarning3 = "\" track, so we'll insert a dummy \"????\" Media Data Atom instead. A separate, codec-specific editing pass will be needed before this track can be played.\n" ; |
580 | |
581 | do { |
582 | fQTEnableTrack = True; // enable this track in the movie by default |
583 | fQTTimeScale = fOurSubsession.rtpTimestampFrequency(); // by default |
584 | fQTTimeUnitsPerSample = 1; // by default |
585 | fQTBytesPerFrame = 0; |
586 | // by default - indicates that the whole packet data is a frame |
587 | fQTSamplesPerFrame = 1; // by default |
588 | |
589 | // Make sure our subsession's medium is one that we know how to |
590 | // represent in a QuickTime file: |
591 | if (isHintTrack()) { |
592 | // Hint tracks are treated specially |
593 | fQTEnableTrack = False; // hint tracks are marked as inactive |
594 | fQTcomponentSubtype = fourChar('h','i','n','t'); |
595 | fQTcomponentName = "hint media handler" ; |
596 | fQTMediaInformationAtomCreator = &QuickTimeFileSink::addAtom_gmhd; |
597 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_rtp; |
598 | } else if (strcmp(fOurSubsession.mediumName(), "audio" ) == 0) { |
599 | fQTcomponentSubtype = fourChar('s','o','u','n'); |
600 | fQTcomponentName = "Apple Sound Media Handler" ; |
601 | fQTMediaInformationAtomCreator = &QuickTimeFileSink::addAtom_smhd; |
602 | fQTMediaDataAtomCreator |
603 | = &QuickTimeFileSink::addAtom_soundMediaGeneral; // by default |
604 | fQTSoundSampleVersion = 0; // by default |
605 | |
606 | // Make sure that our subsession's codec is one that we can handle: |
607 | if (strcmp(fOurSubsession.codecName(), "X-QT" ) == 0 || |
608 | strcmp(fOurSubsession.codecName(), "X-QUICKTIME" ) == 0) { |
609 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_genericMedia; |
610 | } else if (strcmp(fOurSubsession.codecName(), "PCMU" ) == 0) { |
611 | fQTAudioDataType = "ulaw" ; |
612 | fQTBytesPerFrame = 1; |
613 | } else if (strcmp(fOurSubsession.codecName(), "GSM" ) == 0) { |
614 | fQTAudioDataType = "agsm" ; |
615 | fQTBytesPerFrame = 33; |
616 | fQTSamplesPerFrame = 160; |
617 | } else if (strcmp(fOurSubsession.codecName(), "PCMA" ) == 0) { |
618 | fQTAudioDataType = "alaw" ; |
619 | fQTBytesPerFrame = 1; |
620 | } else if (strcmp(fOurSubsession.codecName(), "QCELP" ) == 0) { |
621 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_Qclp; |
622 | fQTSamplesPerFrame = 160; |
623 | } else if (strcmp(fOurSubsession.codecName(), "MPEG4-GENERIC" ) == 0 || |
624 | strcmp(fOurSubsession.codecName(), "MP4A-LATM" ) == 0) { |
625 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_mp4a; |
626 | fQTTimeUnitsPerSample = 1024; // QT considers each frame to be a 'sample' |
627 | // The time scale (frequency) comes from the 'config' information. |
628 | // It might be different from the RTP timestamp frequency (e.g., aacPlus). |
629 | unsigned frequencyFromConfig |
630 | = samplingFrequencyFromAudioSpecificConfig(fOurSubsession.fmtp_config()); |
631 | if (frequencyFromConfig != 0) fQTTimeScale = frequencyFromConfig; |
632 | } else { |
633 | envir() << noCodecWarning1 << "Audio" << noCodecWarning2 |
634 | << fOurSubsession.codecName() << noCodecWarning3; |
635 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_dummy; |
636 | fQTEnableTrack = False; // disable this track in the movie |
637 | } |
638 | } else if (strcmp(fOurSubsession.mediumName(), "video" ) == 0) { |
639 | fQTcomponentSubtype = fourChar('v','i','d','e'); |
640 | fQTcomponentName = "Apple Video Media Handler" ; |
641 | fQTMediaInformationAtomCreator = &QuickTimeFileSink::addAtom_vmhd; |
642 | |
643 | // Make sure that our subsession's codec is one that we can handle: |
644 | if (strcmp(fOurSubsession.codecName(), "X-QT" ) == 0 || |
645 | strcmp(fOurSubsession.codecName(), "X-QUICKTIME" ) == 0) { |
646 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_genericMedia; |
647 | } else if (strcmp(fOurSubsession.codecName(), "H263-1998" ) == 0 || |
648 | strcmp(fOurSubsession.codecName(), "H263-2000" ) == 0) { |
649 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_h263; |
650 | fQTTimeScale = 600; |
651 | fQTTimeUnitsPerSample = fQTTimeScale/fOurSink.fMovieFPS; |
652 | } else if (strcmp(fOurSubsession.codecName(), "H264" ) == 0) { |
653 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_avc1; |
654 | fQTTimeScale = 600; |
655 | fQTTimeUnitsPerSample = fQTTimeScale/fOurSink.fMovieFPS; |
656 | } else if (strcmp(fOurSubsession.codecName(), "MP4V-ES" ) == 0) { |
657 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_mp4v; |
658 | fQTTimeScale = 600; |
659 | fQTTimeUnitsPerSample = fQTTimeScale/fOurSink.fMovieFPS; |
660 | } else { |
661 | envir() << noCodecWarning1 << "Video" << noCodecWarning2 |
662 | << fOurSubsession.codecName() << noCodecWarning3; |
663 | fQTMediaDataAtomCreator = &QuickTimeFileSink::addAtom_dummy; |
664 | fQTEnableTrack = False; // disable this track in the movie |
665 | } |
666 | } else { |
667 | envir() << "Warning: We don't implement a QuickTime Media Handler for media type \"" |
668 | << fOurSubsession.mediumName() << "\"" ; |
669 | break; |
670 | } |
671 | |
672 | #ifdef QT_SUPPORT_PARTIALLY_ONLY |
673 | envir() << "Warning: We don't have sufficient codec-specific information (e.g., sample sizes) to fully generate the \"" |
674 | << fOurSubsession.mediumName() << "/" << fOurSubsession.codecName() |
675 | << "\" track, so we'll disable this track in the movie. A separate, codec-specific editing pass will be needed before this track can be played\n" ; |
676 | fQTEnableTrack = False; // disable this track in the movie |
677 | #endif |
678 | |
679 | return True; |
680 | } while (0); |
681 | |
682 | envir() << ", so a track for the \"" << fOurSubsession.mediumName() |
683 | << "/" << fOurSubsession.codecName() |
684 | << "\" subsession will not be included in the output QuickTime file\n" ; |
685 | return False; |
686 | } |
687 | |
688 | void SubsessionIOState::setFinalQTstate() { |
689 | // Compute derived parameters, by running through the list of chunks: |
690 | fQTDurationT = 0; |
691 | |
692 | ChunkDescriptor* chunk = fHeadChunk; |
693 | while (chunk != NULL) { |
694 | unsigned const numFrames = chunk->fNumFrames; |
695 | unsigned const dur = numFrames*chunk->fFrameDuration; |
696 | fQTDurationT += dur; |
697 | |
698 | chunk = chunk->fNextChunk; |
699 | } |
700 | |
701 | // Convert this duration from track to movie time scale: |
702 | double scaleFactor = fOurSink.movieTimeScale()/(double)fQTTimeScale; |
703 | fQTDurationM = (unsigned)(fQTDurationT*scaleFactor); |
704 | |
705 | if (fQTDurationM > fOurSink.fMaxTrackDurationM) { |
706 | fOurSink.fMaxTrackDurationM = fQTDurationM; |
707 | } |
708 | } |
709 | |
710 | void SubsessionIOState::afterGettingFrame(unsigned packetDataSize, |
711 | struct timeval presentationTime) { |
712 | // Begin by checking whether there was a gap in the RTP stream. |
713 | // If so, try to compensate for this (if desired): |
714 | if (fOurSubsession.rtpSource() != NULL) { // we have a RTP stream |
715 | unsigned short rtpSeqNum |
716 | = fOurSubsession.rtpSource()->curPacketRTPSeqNum(); |
717 | if (fOurSink.fPacketLossCompensate && fPrevBuffer->bytesInUse() > 0) { |
718 | short seqNumGap = rtpSeqNum - fLastPacketRTPSeqNum; |
719 | for (short i = 1; i < seqNumGap; ++i) { |
720 | // Insert a copy of the previous frame, to compensate for the loss: |
721 | useFrame(*fPrevBuffer); |
722 | } |
723 | } |
724 | fLastPacketRTPSeqNum = rtpSeqNum; |
725 | } |
726 | |
727 | // Now, continue working with the frame that we just got |
728 | fOurSink.noteRecordedFrame(fOurSubsession, packetDataSize, presentationTime); |
729 | |
730 | if (fBuffer->bytesInUse() == 0) { |
731 | fBuffer->setPresentationTime(presentationTime); |
732 | } |
733 | fBuffer->addBytes(packetDataSize); |
734 | |
735 | // If our RTP source is a "QuickTimeGenericRTPSource", then |
736 | // use its 'qtState' to set some parameters that we need: |
737 | if (fOurSubsession.rtpSource() != NULL // we have a RTP stream |
738 | && fQTMediaDataAtomCreator == &QuickTimeFileSink::addAtom_genericMedia) { |
739 | QuickTimeGenericRTPSource* rtpSource |
740 | = (QuickTimeGenericRTPSource*)fOurSubsession.rtpSource(); |
741 | QuickTimeGenericRTPSource::QTState& qtState = rtpSource->qtState; |
742 | fQTTimeScale = qtState.timescale; |
743 | if (qtState.width != 0) { |
744 | fOurSink.fMovieWidth = qtState.width; |
745 | } |
746 | if (qtState.height != 0) { |
747 | fOurSink.fMovieHeight = qtState.height; |
748 | } |
749 | |
750 | // Also, if the media type in the "sdAtom" is one that we recognize |
751 | // to have a special parameters, then fix this here: |
752 | if (qtState.sdAtomSize >= 8) { |
753 | char const* atom = qtState.sdAtom; |
754 | unsigned mediaType = fourChar(atom[4],atom[5],atom[6],atom[7]); |
755 | switch (mediaType) { |
756 | case fourChar('a','g','s','m'): { |
757 | fQTBytesPerFrame = 33; |
758 | fQTSamplesPerFrame = 160; |
759 | break; |
760 | } |
761 | case fourChar('Q','c','l','p'): { |
762 | fQTBytesPerFrame = 35; |
763 | fQTSamplesPerFrame = 160; |
764 | break; |
765 | } |
766 | case fourChar('H','c','l','p'): { |
767 | fQTBytesPerFrame = 17; |
768 | fQTSamplesPerFrame = 160; |
769 | break; |
770 | } |
771 | case fourChar('h','2','6','3'): { |
772 | fQTTimeUnitsPerSample = fQTTimeScale/fOurSink.fMovieFPS; |
773 | break; |
774 | } |
775 | } |
776 | } |
777 | } else if (fQTMediaDataAtomCreator == &QuickTimeFileSink::addAtom_Qclp) { |
778 | // For QCELP data, make a note of the frame size (even though it's the |
779 | // same as the packet data size), because it varies depending on the |
780 | // 'rate' of the stream, and this size gets used later when setting up |
781 | // the 'Qclp' QuickTime atom: |
782 | fQTBytesPerFrame = packetDataSize; |
783 | } |
784 | |
785 | useFrame(*fBuffer); |
786 | if (fOurSink.fPacketLossCompensate) { |
787 | // Save this frame, in case we need it for recovery: |
788 | SubsessionBuffer* tmp = fPrevBuffer; // assert: != NULL |
789 | fPrevBuffer = fBuffer; |
790 | fBuffer = tmp; |
791 | } |
792 | fBuffer->reset(); // for the next input |
793 | |
794 | // Now, try getting more frames: |
795 | fOurSink.continuePlaying(); |
796 | } |
797 | |
798 | void SubsessionIOState::useFrame(SubsessionBuffer& buffer) { |
799 | unsigned char* const frameSource = buffer.dataStart(); |
800 | unsigned const frameSize = buffer.bytesInUse(); |
801 | struct timeval const& presentationTime = buffer.presentationTime(); |
802 | int64_t const destFileOffset = TellFile64(fOurSink.fOutFid); |
803 | unsigned sampleNumberOfFrameStart = fQTTotNumSamples + 1; |
804 | Boolean avcHack = fQTMediaDataAtomCreator == &QuickTimeFileSink::addAtom_avc1; |
805 | |
806 | // If we're not syncing streams, or this subsession is not video, then |
807 | // just give this frame a fixed duration: |
808 | if (!fOurSink.fSyncStreams |
809 | || fQTcomponentSubtype != fourChar('v','i','d','e')) { |
810 | unsigned const frameDuration = fQTTimeUnitsPerSample*fQTSamplesPerFrame; |
811 | unsigned frameSizeToUse = frameSize; |
812 | if (avcHack) frameSizeToUse += 4; // H.264/AVC gets the frame size prefix |
813 | |
814 | fQTTotNumSamples += useFrame1(frameSizeToUse, presentationTime, frameDuration, destFileOffset); |
815 | } else { |
816 | // For synced video streams, we use the difference between successive |
817 | // frames' presentation times as the 'frame duration'. So, record |
818 | // information about the *previous* frame: |
819 | struct timeval const& ppt = fPrevFrameState.presentationTime; //abbrev |
820 | if (ppt.tv_sec != 0 || ppt.tv_usec != 0) { |
821 | // There has been a previous frame. |
822 | double duration = (presentationTime.tv_sec - ppt.tv_sec) |
823 | + (presentationTime.tv_usec - ppt.tv_usec)/1000000.0; |
824 | if (duration < 0.0) duration = 0.0; |
825 | unsigned frameDuration |
826 | = (unsigned)((2*duration*fQTTimeScale+1)/2); // round |
827 | unsigned frameSizeToUse = fPrevFrameState.frameSize; |
828 | if (avcHack) frameSizeToUse += 4; // H.264/AVC gets the frame size prefix |
829 | |
830 | unsigned numSamples |
831 | = useFrame1(frameSizeToUse, ppt, frameDuration, fPrevFrameState.destFileOffset); |
832 | fQTTotNumSamples += numSamples; |
833 | sampleNumberOfFrameStart = fQTTotNumSamples + 1; |
834 | } |
835 | |
836 | if (avcHack && (*frameSource == H264_IDR_FRAME)) { |
837 | SyncFrame* newSyncFrame = new SyncFrame(fQTTotNumSamples + 1); |
838 | if (fTailSyncFrame == NULL) { |
839 | fHeadSyncFrame = newSyncFrame; |
840 | } else { |
841 | fTailSyncFrame->nextSyncFrame = newSyncFrame; |
842 | } |
843 | fTailSyncFrame = newSyncFrame; |
844 | } |
845 | |
846 | // Remember the current frame for next time: |
847 | fPrevFrameState.frameSize = frameSize; |
848 | fPrevFrameState.presentationTime = presentationTime; |
849 | fPrevFrameState.destFileOffset = destFileOffset; |
850 | } |
851 | |
852 | if (avcHack) fOurSink.addWord(frameSize); |
853 | |
854 | // Write the data into the file: |
855 | fwrite(frameSource, 1, frameSize, fOurSink.fOutFid); |
856 | |
857 | // If we have a hint track, then write to it also (only if we have a RTP stream): |
858 | if (hasHintTrack() && fOurSubsession.rtpSource() != NULL) { |
859 | // Because presentation times are used for RTP packet timestamps, |
860 | // we don't starting writing to the hint track until we've been synced: |
861 | if (!fHaveBeenSynced) { |
862 | fHaveBeenSynced = fOurSubsession.rtpSource()->hasBeenSynchronizedUsingRTCP(); |
863 | } |
864 | if (fHaveBeenSynced) { |
865 | fHintTrackForUs->useFrameForHinting(frameSize, presentationTime, |
866 | sampleNumberOfFrameStart); |
867 | } |
868 | } |
869 | } |
870 | |
871 | void SubsessionIOState::useFrameForHinting(unsigned frameSize, |
872 | struct timeval presentationTime, |
873 | unsigned startSampleNumber) { |
874 | // At this point, we have a single, combined frame - not individual packets. |
875 | // For the hint track, we need to split the frame back up into separate packets. |
876 | // However, for some RTP sources, then we also need to reuse the special |
877 | // header bytes that were at the start of each of the RTP packets. |
878 | Boolean hack263 = strcmp(fOurSubsession.codecName(), "H263-1998" ) == 0; |
879 | Boolean hackm4a_generic = strcmp(fOurSubsession.mediumName(), "audio" ) == 0 |
880 | && strcmp(fOurSubsession.codecName(), "MPEG4-GENERIC" ) == 0; |
881 | Boolean hackm4a_latm = strcmp(fOurSubsession.mediumName(), "audio" ) == 0 |
882 | && strcmp(fOurSubsession.codecName(), "MP4A-LATM" ) == 0; |
883 | Boolean hackm4a = hackm4a_generic || hackm4a_latm; |
884 | Boolean = (hack263 || hackm4a_generic); |
885 | |
886 | // If there has been a previous frame, then output a 'hint sample' for it. |
887 | // (We use the current frame's presentation time to compute the previous |
888 | // hint sample's duration.) |
889 | RTPSource* const rs = fOurSubsession.rtpSource(); // abbrev (ASSERT: != NULL) |
890 | struct timeval const& ppt = fPrevFrameState.presentationTime; //abbrev |
891 | if (ppt.tv_sec != 0 || ppt.tv_usec != 0) { |
892 | double duration = (presentationTime.tv_sec - ppt.tv_sec) |
893 | + (presentationTime.tv_usec - ppt.tv_usec)/1000000.0; |
894 | if (duration < 0.0) duration = 0.0; |
895 | unsigned msDuration = (unsigned)(duration*1000); // milliseconds |
896 | if (msDuration > fHINF.dmax) fHINF.dmax = msDuration; |
897 | unsigned hintSampleDuration |
898 | = (unsigned)((2*duration*fQTTimeScale+1)/2); // round |
899 | if (hackm4a) { |
900 | // Because multiple AAC frames can appear in a RTP packet, the presentation |
901 | // times of the second and subsequent frames will not be accurate. |
902 | // So, use the known "hintSampleDuration" instead: |
903 | hintSampleDuration = fTrackHintedByUs->fQTTimeUnitsPerSample; |
904 | |
905 | // Also, if the 'time scale' was different from the RTP timestamp frequency, |
906 | // (as can happen with aacPlus), then we need to scale "hintSampleDuration" |
907 | // accordingly: |
908 | if (fTrackHintedByUs->fQTTimeScale != fOurSubsession.rtpTimestampFrequency()) { |
909 | unsigned const scalingFactor |
910 | = fOurSubsession.rtpTimestampFrequency()/fTrackHintedByUs->fQTTimeScale ; |
911 | hintSampleDuration *= scalingFactor; |
912 | } |
913 | } |
914 | |
915 | int64_t const hintSampleDestFileOffset = TellFile64(fOurSink.fOutFid); |
916 | |
917 | unsigned const maxPacketSize = 1450; |
918 | unsigned short numPTEntries |
919 | = (fPrevFrameState.frameSize + (maxPacketSize-1))/maxPacketSize; // normal case |
920 | unsigned char* immediateDataPtr = NULL; |
921 | unsigned immediateDataBytesRemaining = 0; |
922 | if (haveSpecialHeaders) { // special case |
923 | numPTEntries = fPrevFrameState.numSpecialHeaders; |
924 | immediateDataPtr = fPrevFrameState.specialHeaderBytes; |
925 | immediateDataBytesRemaining |
926 | = fPrevFrameState.specialHeaderBytesLength; |
927 | } |
928 | unsigned hintSampleSize |
929 | = fOurSink.addHalfWord(numPTEntries);// Entry count |
930 | hintSampleSize += fOurSink.addHalfWord(0x0000); // Reserved |
931 | |
932 | unsigned offsetWithinSample = 0; |
933 | for (unsigned i = 0; i < numPTEntries; ++i) { |
934 | // Output a Packet Table entry (representing a single RTP packet): |
935 | unsigned short numDTEntries = 1; |
936 | unsigned short seqNum = fPrevFrameState.seqNum++; |
937 | // Note: This assumes that the input stream had no packets lost ##### |
938 | unsigned = fPrevFrameState.rtpHeader; |
939 | if (i+1 < numPTEntries) { |
940 | // This is not the last RTP packet, so clear the marker bit: |
941 | rtpHeader &=~ (1<<23); |
942 | } |
943 | unsigned dataFrameSize = (i+1 < numPTEntries) |
944 | ? maxPacketSize : fPrevFrameState.frameSize - i*maxPacketSize; // normal case |
945 | unsigned sampleNumber = fPrevFrameState.startSampleNumber; |
946 | |
947 | unsigned char immediateDataLen = 0; |
948 | if (haveSpecialHeaders) { // special case |
949 | ++numDTEntries; // to include a Data Table entry for the special hdr |
950 | if (immediateDataBytesRemaining > 0) { |
951 | if (hack263) { |
952 | immediateDataLen = *immediateDataPtr++; |
953 | --immediateDataBytesRemaining; |
954 | if (immediateDataLen > immediateDataBytesRemaining) { |
955 | // shouldn't happen (length byte was bad) |
956 | immediateDataLen = immediateDataBytesRemaining; |
957 | } |
958 | } else { |
959 | immediateDataLen = fPrevFrameState.specialHeaderBytesLength; |
960 | } |
961 | } |
962 | dataFrameSize = fPrevFrameState.packetSizes[i] - immediateDataLen; |
963 | |
964 | if (hack263) { |
965 | Boolean PbitSet |
966 | = immediateDataLen >= 1 && (immediateDataPtr[0]&0x4) != 0; |
967 | if (PbitSet) { |
968 | offsetWithinSample += 2; // to omit the two leading 0 bytes |
969 | } |
970 | } |
971 | } |
972 | |
973 | // Output the Packet Table: |
974 | hintSampleSize += fOurSink.addWord(0); // Relative transmission time |
975 | hintSampleSize += fOurSink.addWord(rtpHeader|seqNum); |
976 | // RTP header info + RTP sequence number |
977 | hintSampleSize += fOurSink.addHalfWord(0x0000); // Flags |
978 | hintSampleSize += fOurSink.addHalfWord(numDTEntries); // Entry count |
979 | unsigned totalPacketSize = 0; |
980 | |
981 | // Output the Data Table: |
982 | if (haveSpecialHeaders) { |
983 | // use the "Immediate Data" format (1): |
984 | hintSampleSize += fOurSink.addByte(1); // Source |
985 | unsigned char len = immediateDataLen > 14 ? 14 : immediateDataLen; |
986 | hintSampleSize += fOurSink.addByte(len); // Length |
987 | totalPacketSize += len; fHINF.dimm += len; |
988 | unsigned char j; |
989 | for (j = 0; j < len; ++j) { |
990 | hintSampleSize += fOurSink.addByte(immediateDataPtr[j]); // Data |
991 | } |
992 | for (j = len; j < 14; ++j) { |
993 | hintSampleSize += fOurSink.addByte(0); // Data (padding) |
994 | } |
995 | |
996 | immediateDataPtr += immediateDataLen; |
997 | immediateDataBytesRemaining -= immediateDataLen; |
998 | } |
999 | // use the "Sample Data" format (2): |
1000 | hintSampleSize += fOurSink.addByte(2); // Source |
1001 | hintSampleSize += fOurSink.addByte(0); // Track ref index |
1002 | hintSampleSize += fOurSink.addHalfWord(dataFrameSize); // Length |
1003 | totalPacketSize += dataFrameSize; fHINF.dmed += dataFrameSize; |
1004 | hintSampleSize += fOurSink.addWord(sampleNumber); // Sample number |
1005 | hintSampleSize += fOurSink.addWord(offsetWithinSample); // Offset |
1006 | // Get "bytes|samples per compression block" from the hinted track: |
1007 | unsigned short const bytesPerCompressionBlock |
1008 | = fTrackHintedByUs->fQTBytesPerFrame; |
1009 | unsigned short const samplesPerCompressionBlock |
1010 | = fTrackHintedByUs->fQTSamplesPerFrame; |
1011 | hintSampleSize += fOurSink.addHalfWord(bytesPerCompressionBlock); |
1012 | hintSampleSize += fOurSink.addHalfWord(samplesPerCompressionBlock); |
1013 | |
1014 | offsetWithinSample += dataFrameSize;// for the next iteration (if any) |
1015 | |
1016 | // Tally statistics for this packet: |
1017 | fHINF.nump += 1; |
1018 | fHINF.tpyl += totalPacketSize; |
1019 | totalPacketSize += 12; // add in the size of the RTP header |
1020 | fHINF.trpy += totalPacketSize; |
1021 | if (totalPacketSize > fHINF.pmax) fHINF.pmax = totalPacketSize; |
1022 | } |
1023 | |
1024 | // Make note of this completed hint sample frame: |
1025 | fQTTotNumSamples += useFrame1(hintSampleSize, ppt, hintSampleDuration, |
1026 | hintSampleDestFileOffset); |
1027 | } |
1028 | |
1029 | // Remember this frame for next time: |
1030 | fPrevFrameState.frameSize = frameSize; |
1031 | fPrevFrameState.presentationTime = presentationTime; |
1032 | fPrevFrameState.startSampleNumber = startSampleNumber; |
1033 | fPrevFrameState.rtpHeader |
1034 | = rs->curPacketMarkerBit()<<23 |
1035 | | (rs->rtpPayloadFormat()&0x7F)<<16; |
1036 | if (hack263) { |
1037 | H263plusVideoRTPSource* rs_263 = (H263plusVideoRTPSource*)rs; |
1038 | fPrevFrameState.numSpecialHeaders = rs_263->fNumSpecialHeaders; |
1039 | fPrevFrameState.specialHeaderBytesLength = rs_263->fSpecialHeaderBytesLength; |
1040 | unsigned i; |
1041 | for (i = 0; i < rs_263->fSpecialHeaderBytesLength; ++i) { |
1042 | fPrevFrameState.specialHeaderBytes[i] = rs_263->fSpecialHeaderBytes[i]; |
1043 | } |
1044 | for (i = 0; i < rs_263->fNumSpecialHeaders; ++i) { |
1045 | fPrevFrameState.packetSizes[i] = rs_263->fPacketSizes[i]; |
1046 | } |
1047 | } else if (hackm4a_generic) { |
1048 | // Synthesize a special header, so that this frame can be in its own RTP packet. |
1049 | unsigned const sizeLength = fOurSubsession.attrVal_unsigned("sizelength" ); |
1050 | unsigned const indexLength = fOurSubsession.attrVal_unsigned("indexlength" ); |
1051 | if (sizeLength + indexLength != 16) { |
1052 | envir() << "Warning: unexpected 'sizeLength' " << sizeLength |
1053 | << " and 'indexLength' " << indexLength |
1054 | << "seen when creating hint track\n" ; |
1055 | } |
1056 | fPrevFrameState.numSpecialHeaders = 1; |
1057 | fPrevFrameState.specialHeaderBytesLength = 4; |
1058 | fPrevFrameState.specialHeaderBytes[0] = 0; // AU_headers_length (high byte) |
1059 | fPrevFrameState.specialHeaderBytes[1] = 16; // AU_headers_length (low byte) |
1060 | fPrevFrameState.specialHeaderBytes[2] = ((frameSize<<indexLength)&0xFF00)>>8; |
1061 | fPrevFrameState.specialHeaderBytes[3] = (frameSize<<indexLength); |
1062 | fPrevFrameState.packetSizes[0] |
1063 | = fPrevFrameState.specialHeaderBytesLength + frameSize; |
1064 | } |
1065 | } |
1066 | |
1067 | unsigned SubsessionIOState::useFrame1(unsigned sourceDataSize, |
1068 | struct timeval presentationTime, |
1069 | unsigned frameDuration, |
1070 | int64_t destFileOffset) { |
1071 | // Figure out the actual frame size for this data: |
1072 | unsigned frameSize = fQTBytesPerFrame; |
1073 | if (frameSize == 0) { |
1074 | // The entire packet data is assumed to be a frame: |
1075 | frameSize = sourceDataSize; |
1076 | } |
1077 | unsigned const numFrames = sourceDataSize/frameSize; |
1078 | unsigned const numSamples = numFrames*fQTSamplesPerFrame; |
1079 | |
1080 | // Record the information about which 'chunk' this data belongs to: |
1081 | ChunkDescriptor* newTailChunk; |
1082 | if (fTailChunk == NULL) { |
1083 | newTailChunk = fHeadChunk |
1084 | = new ChunkDescriptor(destFileOffset, sourceDataSize, |
1085 | frameSize, frameDuration, presentationTime); |
1086 | } else { |
1087 | newTailChunk = fTailChunk->extendChunk(destFileOffset, sourceDataSize, |
1088 | frameSize, frameDuration, |
1089 | presentationTime); |
1090 | } |
1091 | if (newTailChunk != fTailChunk) { |
1092 | // This data created a new chunk, rather than extending the old one |
1093 | ++fNumChunks; |
1094 | fTailChunk = newTailChunk; |
1095 | } |
1096 | |
1097 | return numSamples; |
1098 | } |
1099 | |
1100 | void SubsessionIOState::onSourceClosure() { |
1101 | fOurSourceIsActive = False; |
1102 | fOurSink.onSourceClosure1(); |
1103 | } |
1104 | |
1105 | Boolean SubsessionIOState::syncOK(struct timeval presentationTime) { |
1106 | QuickTimeFileSink& s = fOurSink; // abbreviation |
1107 | if (!s.fSyncStreams || fOurSubsession.rtpSource() == NULL) return True; // we don't care |
1108 | |
1109 | if (s.fNumSyncedSubsessions < s.fNumSubsessions) { |
1110 | // Not all subsessions have yet been synced. Check whether ours was |
1111 | // one of the unsynced ones, and, if so, whether it is now synced: |
1112 | if (!fHaveBeenSynced) { |
1113 | // We weren't synchronized before |
1114 | if (fOurSubsession.rtpSource()->hasBeenSynchronizedUsingRTCP()) { |
1115 | // H264 ? |
1116 | if (fQTMediaDataAtomCreator == &QuickTimeFileSink::addAtom_avc1) { |
1117 | // special case: audio + H264 video: wait until audio is in sync |
1118 | if ((s.fNumSubsessions == 2) && (s.fNumSyncedSubsessions < (s.fNumSubsessions - 1))) return False; |
1119 | |
1120 | // if audio is in sync, wait for the next IDR frame to start |
1121 | unsigned char* const frameSource = fBuffer->dataStart(); |
1122 | if (*frameSource != H264_IDR_FRAME) return False; |
1123 | } |
1124 | // But now we are |
1125 | fHaveBeenSynced = True; |
1126 | fSyncTime = presentationTime; |
1127 | ++s.fNumSyncedSubsessions; |
1128 | |
1129 | if (timevalGE(fSyncTime, s.fNewestSyncTime)) { |
1130 | s.fNewestSyncTime = fSyncTime; |
1131 | } |
1132 | } |
1133 | } |
1134 | } |
1135 | |
1136 | // Check again whether all subsessions have been synced: |
1137 | if (s.fNumSyncedSubsessions < s.fNumSubsessions) return False; |
1138 | |
1139 | // Allow this data if it is more recent than the newest sync time: |
1140 | return timevalGE(presentationTime, s.fNewestSyncTime); |
1141 | } |
1142 | |
1143 | void SubsessionIOState::setHintTrack(SubsessionIOState* hintedTrack, |
1144 | SubsessionIOState* hintTrack) { |
1145 | if (hintedTrack != NULL) hintedTrack->fHintTrackForUs = hintTrack; |
1146 | if (hintTrack != NULL) hintTrack->fTrackHintedByUs = hintedTrack; |
1147 | } |
1148 | |
1149 | SyncFrame::SyncFrame(unsigned ) |
1150 | : nextSyncFrame(NULL), sfFrameNum(frameNum) { |
1151 | } |
1152 | |
1153 | void Count64::operator+=(unsigned arg) { |
1154 | unsigned newLo = lo + arg; |
1155 | if (newLo < lo) { // lo has overflowed |
1156 | ++hi; |
1157 | } |
1158 | lo = newLo; |
1159 | } |
1160 | |
1161 | ChunkDescriptor |
1162 | ::ChunkDescriptor(int64_t offsetInFile, unsigned size, |
1163 | unsigned frameSize, unsigned frameDuration, |
1164 | struct timeval presentationTime) |
1165 | : fNextChunk(NULL), fOffsetInFile(offsetInFile), |
1166 | fNumFrames(size/frameSize), |
1167 | fFrameSize(frameSize), fFrameDuration(frameDuration), |
1168 | fPresentationTime(presentationTime) { |
1169 | } |
1170 | |
1171 | ChunkDescriptor* ChunkDescriptor |
1172 | ::extendChunk(int64_t newOffsetInFile, unsigned newSize, |
1173 | unsigned newFrameSize, unsigned newFrameDuration, |
1174 | struct timeval newPresentationTime) { |
1175 | // First, check whether the new space is just at the end of this |
1176 | // existing chunk: |
1177 | if (newOffsetInFile == fOffsetInFile + fNumFrames*fFrameSize) { |
1178 | // We can extend this existing chunk, provided that the frame size |
1179 | // and frame duration have not changed: |
1180 | if (newFrameSize == fFrameSize && newFrameDuration == fFrameDuration) { |
1181 | fNumFrames += newSize/fFrameSize; |
1182 | return this; |
1183 | } |
1184 | } |
1185 | |
1186 | // We'll allocate a new ChunkDescriptor, and link it to the end of us: |
1187 | ChunkDescriptor* newDescriptor |
1188 | = new ChunkDescriptor(newOffsetInFile, newSize, |
1189 | newFrameSize, newFrameDuration, |
1190 | newPresentationTime); |
1191 | |
1192 | fNextChunk = newDescriptor; |
1193 | |
1194 | return newDescriptor; |
1195 | } |
1196 | |
1197 | |
1198 | ////////// QuickTime-specific implementation ////////// |
1199 | |
1200 | unsigned QuickTimeFileSink::addWord64(u_int64_t word) { |
1201 | addByte((unsigned char)(word>>56)); addByte((unsigned char)(word>>48)); |
1202 | addByte((unsigned char)(word>>40)); addByte((unsigned char)(word>>32)); |
1203 | addByte((unsigned char)(word>>24)); addByte((unsigned char)(word>>16)); |
1204 | addByte((unsigned char)(word>>8)); addByte((unsigned char)(word)); |
1205 | |
1206 | return 8; |
1207 | } |
1208 | |
1209 | unsigned QuickTimeFileSink::addWord(unsigned word) { |
1210 | addByte(word>>24); addByte(word>>16); |
1211 | addByte(word>>8); addByte(word); |
1212 | |
1213 | return 4; |
1214 | } |
1215 | |
1216 | unsigned QuickTimeFileSink::addHalfWord(unsigned short halfWord) { |
1217 | addByte((unsigned char)(halfWord>>8)); addByte((unsigned char)halfWord); |
1218 | |
1219 | return 2; |
1220 | } |
1221 | |
1222 | unsigned QuickTimeFileSink::addZeroWords(unsigned numWords) { |
1223 | for (unsigned i = 0; i < numWords; ++i) { |
1224 | addWord(0); |
1225 | } |
1226 | |
1227 | return numWords*4; |
1228 | } |
1229 | |
1230 | unsigned QuickTimeFileSink::add4ByteString(char const* str) { |
1231 | addByte(str[0]); addByte(str[1]); addByte(str[2]); addByte(str[3]); |
1232 | |
1233 | return 4; |
1234 | } |
1235 | |
1236 | unsigned QuickTimeFileSink::addArbitraryString(char const* str, |
1237 | Boolean oneByteLength) { |
1238 | unsigned size = 0; |
1239 | if (oneByteLength) { |
1240 | // Begin with a byte containing the string length: |
1241 | unsigned strLength = strlen(str); |
1242 | if (strLength >= 256) { |
1243 | envir() << "QuickTimeFileSink::addArbitraryString(\"" |
1244 | << str << "\") saw string longer than we know how to handle (" |
1245 | << strLength << "\n" ; |
1246 | } |
1247 | size += addByte((unsigned char)strLength); |
1248 | } |
1249 | |
1250 | while (*str != '\0') { |
1251 | size += addByte(*str++); |
1252 | } |
1253 | |
1254 | return size; |
1255 | } |
1256 | |
1257 | unsigned QuickTimeFileSink::(char const* atomName) { |
1258 | // Output a placeholder for the 4-byte size: |
1259 | addWord(0); |
1260 | |
1261 | // Output the 4-byte atom name: |
1262 | add4ByteString(atomName); |
1263 | |
1264 | return 8; |
1265 | } |
1266 | |
1267 | unsigned QuickTimeFileSink::(char const* atomName) { |
1268 | // Output 64Bit size marker |
1269 | addWord(1); |
1270 | |
1271 | // Output the 4-byte atom name: |
1272 | add4ByteString(atomName); |
1273 | |
1274 | addWord64(0); |
1275 | |
1276 | return 16; |
1277 | } |
1278 | |
1279 | void QuickTimeFileSink::setWord(int64_t filePosn, unsigned size) { |
1280 | do { |
1281 | if (SeekFile64(fOutFid, filePosn, SEEK_SET) < 0) break; |
1282 | addWord(size); |
1283 | if (SeekFile64(fOutFid, 0, SEEK_END) < 0) break; // go back to where we were |
1284 | |
1285 | return; |
1286 | } while (0); |
1287 | |
1288 | // One of the SeekFile64()s failed, probable because we're not a seekable file |
1289 | envir() << "QuickTimeFileSink::setWord(): SeekFile64 failed (err " |
1290 | << envir().getErrno() << ")\n" ; |
1291 | } |
1292 | |
1293 | void QuickTimeFileSink::setWord64(int64_t filePosn, u_int64_t size) { |
1294 | do { |
1295 | if (SeekFile64(fOutFid, filePosn, SEEK_SET) < 0) break; |
1296 | addWord64(size); |
1297 | if (SeekFile64(fOutFid, 0, SEEK_END) < 0) break; // go back to where we were |
1298 | |
1299 | return; |
1300 | } while (0); |
1301 | |
1302 | // One of the SeekFile64()s failed, probable because we're not a seekable file |
1303 | envir() << "QuickTimeFileSink::setWord64(): SeekFile64 failed (err " |
1304 | << envir().getErrno() << ")\n" ; |
1305 | } |
1306 | |
1307 | // Methods for writing particular atoms. Note the following macros: |
1308 | |
1309 | #define addAtom(name) \ |
1310 | unsigned QuickTimeFileSink::addAtom_##name() { \ |
1311 | int64_t initFilePosn = TellFile64(fOutFid); \ |
1312 | unsigned size = addAtomHeader("" #name "") |
1313 | |
1314 | #define addAtomEnd \ |
1315 | setWord(initFilePosn, size); \ |
1316 | return size; \ |
1317 | } |
1318 | |
1319 | addAtom(ftyp); |
1320 | size += add4ByteString("mp42" ); |
1321 | size += addWord(0x00000000); |
1322 | size += add4ByteString("mp42" ); |
1323 | size += add4ByteString("isom" ); |
1324 | addAtomEnd; |
1325 | |
1326 | addAtom(moov); |
1327 | size += addAtom_mvhd(); |
1328 | |
1329 | if (fGenerateMP4Format) { |
1330 | size += addAtom_iods(); |
1331 | } |
1332 | |
1333 | // Add a 'trak' atom for each subsession: |
1334 | // (For some unknown reason, QuickTime Player (5.0 at least) |
1335 | // doesn't display the movie correctly unless the audio track |
1336 | // (if present) appears before the video track. So ensure this here.) |
1337 | MediaSubsessionIterator iter(fInputSession); |
1338 | MediaSubsession* subsession; |
1339 | while ((subsession = iter.next()) != NULL) { |
1340 | fCurrentIOState = (SubsessionIOState*)(subsession->miscPtr); |
1341 | if (fCurrentIOState == NULL) continue; |
1342 | if (strcmp(subsession->mediumName(), "audio" ) != 0) continue; |
1343 | |
1344 | size += addAtom_trak(); |
1345 | |
1346 | if (fCurrentIOState->hasHintTrack()) { |
1347 | // This track has a hint track; output it also: |
1348 | fCurrentIOState = fCurrentIOState->fHintTrackForUs; |
1349 | size += addAtom_trak(); |
1350 | } |
1351 | } |
1352 | iter.reset(); |
1353 | while ((subsession = iter.next()) != NULL) { |
1354 | fCurrentIOState = (SubsessionIOState*)(subsession->miscPtr); |
1355 | if (fCurrentIOState == NULL) continue; |
1356 | if (strcmp(subsession->mediumName(), "audio" ) == 0) continue; |
1357 | |
1358 | size += addAtom_trak(); |
1359 | |
1360 | if (fCurrentIOState->hasHintTrack()) { |
1361 | // This track has a hint track; output it also: |
1362 | fCurrentIOState = fCurrentIOState->fHintTrackForUs; |
1363 | size += addAtom_trak(); |
1364 | } |
1365 | } |
1366 | addAtomEnd; |
1367 | |
1368 | addAtom(mvhd); |
1369 | size += addWord(0x00000000); // Version + Flags |
1370 | size += addWord(fAppleCreationTime); // Creation time |
1371 | size += addWord(fAppleCreationTime); // Modification time |
1372 | |
1373 | // For the "Time scale" field, use the largest RTP timestamp frequency |
1374 | // that we saw in any of the subsessions. |
1375 | size += addWord(movieTimeScale()); // Time scale |
1376 | |
1377 | unsigned const duration = fMaxTrackDurationM; |
1378 | fMVHD_durationPosn = TellFile64(fOutFid); |
1379 | size += addWord(duration); // Duration |
1380 | |
1381 | size += addWord(0x00010000); // Preferred rate |
1382 | size += addWord(0x01000000); // Preferred volume + Reserved[0] |
1383 | size += addZeroWords(2); // Reserved[1-2] |
1384 | size += addWord(0x00010000); // matrix top left corner |
1385 | size += addZeroWords(3); // matrix |
1386 | size += addWord(0x00010000); // matrix center |
1387 | size += addZeroWords(3); // matrix |
1388 | size += addWord(0x40000000); // matrix bottom right corner |
1389 | size += addZeroWords(6); // various time fields |
1390 | size += addWord(SubsessionIOState::fCurrentTrackNumber+1);// Next track ID |
1391 | addAtomEnd; |
1392 | |
1393 | addAtom(iods); |
1394 | size += addWord(0x00000000); // Version + Flags |
1395 | size += addWord(0x10808080); |
1396 | size += addWord(0x07004FFF); |
1397 | size += addWord(0xFF0FFFFF); |
1398 | addAtomEnd; |
1399 | |
1400 | addAtom(trak); |
1401 | size += addAtom_tkhd(); |
1402 | |
1403 | // If we're synchronizing the media streams (or are a hint track), |
1404 | // add an edit list that helps do this: |
1405 | if (fCurrentIOState->fHeadChunk != NULL |
1406 | && (fSyncStreams || fCurrentIOState->isHintTrack())) { |
1407 | size += addAtom_edts(); |
1408 | } |
1409 | |
1410 | // If we're generating a hint track, add a 'tref' atom: |
1411 | if (fCurrentIOState->isHintTrack()) size += addAtom_tref(); |
1412 | |
1413 | size += addAtom_mdia(); |
1414 | |
1415 | // If we're generating a hint track, add a 'udta' atom: |
1416 | if (fCurrentIOState->isHintTrack()) size += addAtom_udta(); |
1417 | addAtomEnd; |
1418 | |
1419 | addAtom(tkhd); |
1420 | if (fCurrentIOState->fQTEnableTrack) { |
1421 | size += addWord(0x0000000F); // Version + Flags |
1422 | } else { |
1423 | // Disable this track in the movie: |
1424 | size += addWord(0x00000000); // Version + Flags |
1425 | } |
1426 | size += addWord(fAppleCreationTime); // Creation time |
1427 | size += addWord(fAppleCreationTime); // Modification time |
1428 | size += addWord(fCurrentIOState->fTrackID); // Track ID |
1429 | size += addWord(0x00000000); // Reserved |
1430 | |
1431 | unsigned const duration = fCurrentIOState->fQTDurationM; // movie units |
1432 | fCurrentIOState->fTKHD_durationPosn = TellFile64(fOutFid); |
1433 | size += addWord(duration); // Duration |
1434 | size += addZeroWords(3); // Reserved+Layer+Alternate grp |
1435 | size += addWord(0x01000000); // Volume + Reserved |
1436 | size += addWord(0x00010000); // matrix top left corner |
1437 | size += addZeroWords(3); // matrix |
1438 | size += addWord(0x00010000); // matrix center |
1439 | size += addZeroWords(3); // matrix |
1440 | size += addWord(0x40000000); // matrix bottom right corner |
1441 | if (strcmp(fCurrentIOState->fOurSubsession.mediumName(), "video" ) == 0) { |
1442 | size += addWord(fMovieWidth<<16); // Track width |
1443 | size += addWord(fMovieHeight<<16); // Track height |
1444 | } else { |
1445 | size += addZeroWords(2); // not video: leave width and height fields zero |
1446 | } |
1447 | addAtomEnd; |
1448 | |
1449 | addAtom(edts); |
1450 | size += addAtom_elst(); |
1451 | addAtomEnd; |
1452 | |
1453 | #define addEdit1(duration,trackPosition) do { \ |
1454 | unsigned trackDuration \ |
1455 | = (unsigned) ((2*(duration)*movieTimeScale()+1)/2); \ |
1456 | /* in movie time units */ \ |
1457 | size += addWord(trackDuration); /* Track duration */ \ |
1458 | totalDurationOfEdits += trackDuration; \ |
1459 | size += addWord(trackPosition); /* Media time */ \ |
1460 | size += addWord(0x00010000); /* Media rate (1x) */ \ |
1461 | ++numEdits; \ |
1462 | } while (0) |
1463 | #define addEdit(duration) addEdit1((duration),editTrackPosition) |
1464 | #define addEmptyEdit(duration) addEdit1((duration),(~0)) |
1465 | |
1466 | addAtom(elst); |
1467 | size += addWord(0x00000000); // Version + Flags |
1468 | |
1469 | // Add a dummy "Number of entries" field |
1470 | // (and remember its position). We'll fill this field in later: |
1471 | int64_t numEntriesPosition = TellFile64(fOutFid); |
1472 | size += addWord(0); // dummy for "Number of entries" |
1473 | unsigned numEdits = 0; |
1474 | unsigned totalDurationOfEdits = 0; // in movie time units |
1475 | |
1476 | // Run through our chunks, looking at their presentation times. |
1477 | // From these, figure out the edits that need to be made to keep |
1478 | // the track media data in sync with the presentation times. |
1479 | |
1480 | double const syncThreshold = 0.1; // 100 ms |
1481 | // don't allow the track to get out of sync by more than this |
1482 | |
1483 | struct timeval editStartTime = fFirstDataTime; |
1484 | unsigned editTrackPosition = 0; |
1485 | unsigned currentTrackPosition = 0; |
1486 | double trackDurationOfEdit = 0.0; |
1487 | unsigned chunkDuration = 0; |
1488 | |
1489 | ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk; |
1490 | while (chunk != NULL) { |
1491 | struct timeval const& chunkStartTime = chunk->fPresentationTime; |
1492 | double movieDurationOfEdit |
1493 | = (chunkStartTime.tv_sec - editStartTime.tv_sec) |
1494 | + (chunkStartTime.tv_usec - editStartTime.tv_usec)/1000000.0; |
1495 | trackDurationOfEdit = (currentTrackPosition-editTrackPosition) |
1496 | / (double)(fCurrentIOState->fQTTimeScale); |
1497 | |
1498 | double outOfSync = movieDurationOfEdit - trackDurationOfEdit; |
1499 | |
1500 | if (outOfSync > syncThreshold) { |
1501 | // The track's data is too short, so end this edit, add a new |
1502 | // 'empty' edit after it, and start a new edit |
1503 | // (at the current track posn.): |
1504 | if (trackDurationOfEdit > 0.0) addEdit(trackDurationOfEdit); |
1505 | addEmptyEdit(outOfSync); |
1506 | |
1507 | editStartTime = chunkStartTime; |
1508 | editTrackPosition = currentTrackPosition; |
1509 | } else if (outOfSync < -syncThreshold) { |
1510 | // The track's data is too long, so end this edit, and start |
1511 | // a new edit (pointing at the current track posn.): |
1512 | if (movieDurationOfEdit > 0.0) addEdit(movieDurationOfEdit); |
1513 | |
1514 | editStartTime = chunkStartTime; |
1515 | editTrackPosition = currentTrackPosition; |
1516 | } |
1517 | |
1518 | // Note the duration of this chunk: |
1519 | unsigned numChannels = fCurrentIOState->fOurSubsession.numChannels(); |
1520 | chunkDuration = chunk->fNumFrames*chunk->fFrameDuration/numChannels; |
1521 | currentTrackPosition += chunkDuration; |
1522 | |
1523 | chunk = chunk->fNextChunk; |
1524 | } |
1525 | |
1526 | // Write out the final edit |
1527 | trackDurationOfEdit |
1528 | += (double)chunkDuration/fCurrentIOState->fQTTimeScale; |
1529 | if (trackDurationOfEdit > 0.0) addEdit(trackDurationOfEdit); |
1530 | |
1531 | // Now go back and fill in the "Number of entries" field: |
1532 | setWord(numEntriesPosition, numEdits); |
1533 | |
1534 | // Also, if the sum of all of the edit durations exceeds the |
1535 | // track duration that we already computed (from sample durations), |
1536 | // then reset the track duration to this new value: |
1537 | if (totalDurationOfEdits > fCurrentIOState->fQTDurationM) { |
1538 | fCurrentIOState->fQTDurationM = totalDurationOfEdits; |
1539 | setWord(fCurrentIOState->fTKHD_durationPosn, totalDurationOfEdits); |
1540 | |
1541 | // Also, check whether the overall movie duration needs to change: |
1542 | if (totalDurationOfEdits > fMaxTrackDurationM) { |
1543 | fMaxTrackDurationM = totalDurationOfEdits; |
1544 | setWord(fMVHD_durationPosn, totalDurationOfEdits); |
1545 | } |
1546 | |
1547 | // Also, convert to track time scale: |
1548 | double scaleFactor |
1549 | = fCurrentIOState->fQTTimeScale/(double)movieTimeScale(); |
1550 | fCurrentIOState->fQTDurationT |
1551 | = (unsigned)(totalDurationOfEdits*scaleFactor); |
1552 | } |
1553 | addAtomEnd; |
1554 | |
1555 | addAtom(tref); |
1556 | size += addAtom_hint(); |
1557 | addAtomEnd; |
1558 | |
1559 | addAtom(hint); |
1560 | SubsessionIOState* hintedTrack = fCurrentIOState->fTrackHintedByUs; |
1561 | // Assert: hintedTrack != NULL |
1562 | size += addWord(hintedTrack->fTrackID); |
1563 | addAtomEnd; |
1564 | |
1565 | addAtom(mdia); |
1566 | size += addAtom_mdhd(); |
1567 | size += addAtom_hdlr(); |
1568 | size += addAtom_minf(); |
1569 | addAtomEnd; |
1570 | |
1571 | addAtom(mdhd); |
1572 | size += addWord(0x00000000); // Version + Flags |
1573 | size += addWord(fAppleCreationTime); // Creation time |
1574 | size += addWord(fAppleCreationTime); // Modification time |
1575 | |
1576 | unsigned const timeScale = fCurrentIOState->fQTTimeScale; |
1577 | size += addWord(timeScale); // Time scale |
1578 | |
1579 | unsigned const duration = fCurrentIOState->fQTDurationT; // track units |
1580 | size += addWord(duration); // Duration |
1581 | |
1582 | size += addWord(0x00000000); // Language+Quality |
1583 | addAtomEnd; |
1584 | |
1585 | addAtom(hdlr); |
1586 | size += addWord(0x00000000); // Version + Flags |
1587 | size += add4ByteString("mhlr" ); // Component type |
1588 | size += addWord(fCurrentIOState->fQTcomponentSubtype); |
1589 | // Component subtype |
1590 | size += add4ByteString("appl" ); // Component manufacturer |
1591 | size += addWord(0x00000000); // Component flags |
1592 | size += addWord(0x00000000); // Component flags mask |
1593 | size += addArbitraryString(fCurrentIOState->fQTcomponentName); |
1594 | // Component name |
1595 | addAtomEnd; |
1596 | |
1597 | addAtom(minf); |
1598 | SubsessionIOState::atomCreationFunc mediaInformationAtomCreator |
1599 | = fCurrentIOState->fQTMediaInformationAtomCreator; |
1600 | size += (this->*mediaInformationAtomCreator)(); |
1601 | size += addAtom_hdlr2(); |
1602 | size += addAtom_dinf(); |
1603 | size += addAtom_stbl(); |
1604 | addAtomEnd; |
1605 | |
1606 | addAtom(smhd); |
1607 | size += addZeroWords(2); // Version+Flags+Balance+Reserved |
1608 | addAtomEnd; |
1609 | |
1610 | addAtom(vmhd); |
1611 | size += addWord(0x00000001); // Version + Flags |
1612 | size += addWord(0x00408000); // Graphics mode + Opcolor[red] |
1613 | size += addWord(0x80008000); // Opcolor[green} + Opcolor[blue] |
1614 | addAtomEnd; |
1615 | |
1616 | addAtom(gmhd); |
1617 | size += addAtom_gmin(); |
1618 | addAtomEnd; |
1619 | |
1620 | addAtom(gmin); |
1621 | size += addWord(0x00000000); // Version + Flags |
1622 | // The following fields probably aren't used for hint tracks, so just |
1623 | // use values that I've seen in other files: |
1624 | size += addWord(0x00408000); // Graphics mode + Opcolor (1st 2 bytes) |
1625 | size += addWord(0x80008000); // Opcolor (last 4 bytes) |
1626 | size += addWord(0x00000000); // Balance + Reserved |
1627 | addAtomEnd; |
1628 | |
1629 | unsigned QuickTimeFileSink::addAtom_hdlr2() { |
1630 | int64_t initFilePosn = TellFile64(fOutFid); |
1631 | unsigned size = addAtomHeader("hdlr" ); |
1632 | size += addWord(0x00000000); // Version + Flags |
1633 | size += add4ByteString("dhlr" ); // Component type |
1634 | size += add4ByteString("alis" ); // Component subtype |
1635 | size += add4ByteString("appl" ); // Component manufacturer |
1636 | size += addZeroWords(2); // Component flags+Component flags mask |
1637 | size += addArbitraryString("Apple Alias Data Handler" ); // Component name |
1638 | addAtomEnd; |
1639 | |
1640 | addAtom(dinf); |
1641 | size += addAtom_dref(); |
1642 | addAtomEnd; |
1643 | |
1644 | addAtom(dref); |
1645 | size += addWord(0x00000000); // Version + Flags |
1646 | size += addWord(0x00000001); // Number of entries |
1647 | size += addAtom_alis(); |
1648 | addAtomEnd; |
1649 | |
1650 | addAtom(alis); |
1651 | size += addWord(0x00000001); // Version + Flags |
1652 | addAtomEnd; |
1653 | |
1654 | addAtom(stbl); |
1655 | size += addAtom_stsd(); |
1656 | size += addAtom_stts(); |
1657 | if (fCurrentIOState->fQTcomponentSubtype == fourChar('v','i','d','e')) { |
1658 | size += addAtom_stss(); // only for video streams |
1659 | } |
1660 | size += addAtom_stsc(); |
1661 | size += addAtom_stsz(); |
1662 | size += addAtom_co64(); |
1663 | addAtomEnd; |
1664 | |
1665 | addAtom(stsd); |
1666 | size += addWord(0x00000000); // Version+Flags |
1667 | size += addWord(0x00000001); // Number of entries |
1668 | SubsessionIOState::atomCreationFunc mediaDataAtomCreator |
1669 | = fCurrentIOState->fQTMediaDataAtomCreator; |
1670 | size += (this->*mediaDataAtomCreator)(); |
1671 | addAtomEnd; |
1672 | |
1673 | unsigned QuickTimeFileSink::addAtom_genericMedia() { |
1674 | int64_t initFilePosn = TellFile64(fOutFid); |
1675 | |
1676 | // Our source is assumed to be a "QuickTimeGenericRTPSource" |
1677 | // Use its "sdAtom" state for our contents: |
1678 | QuickTimeGenericRTPSource* rtpSource = (QuickTimeGenericRTPSource*) |
1679 | fCurrentIOState->fOurSubsession.rtpSource(); |
1680 | unsigned size = 0; |
1681 | if (rtpSource != NULL) { |
1682 | QuickTimeGenericRTPSource::QTState& qtState = rtpSource->qtState; |
1683 | char const* from = qtState.sdAtom; |
1684 | size = qtState.sdAtomSize; |
1685 | for (unsigned i = 0; i < size; ++i) addByte(from[i]); |
1686 | } |
1687 | addAtomEnd; |
1688 | |
1689 | unsigned QuickTimeFileSink::addAtom_soundMediaGeneral() { |
1690 | int64_t initFilePosn = TellFile64(fOutFid); |
1691 | unsigned size = addAtomHeader(fCurrentIOState->fQTAudioDataType); |
1692 | |
1693 | // General sample description fields: |
1694 | size += addWord(0x00000000); // Reserved |
1695 | size += addWord(0x00000001); // Reserved+Data reference index |
1696 | // Sound sample description fields: |
1697 | unsigned short const version = fCurrentIOState->fQTSoundSampleVersion; |
1698 | size += addWord(version<<16); // Version+Revision level |
1699 | size += addWord(0x00000000); // Vendor |
1700 | unsigned short numChannels |
1701 | = (unsigned short)(fCurrentIOState->fOurSubsession.numChannels()); |
1702 | size += addHalfWord(numChannels); // Number of channels |
1703 | size += addHalfWord(0x0010); // Sample size |
1704 | // size += addWord(0x00000000); // Compression ID+Packet size |
1705 | size += addWord(0xfffe0000); // Compression ID+Packet size ##### |
1706 | |
1707 | unsigned const sampleRateFixedPoint = fCurrentIOState->fQTTimeScale << 16; |
1708 | size += addWord(sampleRateFixedPoint); // Sample rate |
1709 | addAtomEnd; |
1710 | |
1711 | unsigned QuickTimeFileSink::addAtom_Qclp() { |
1712 | // The beginning of this atom looks just like a general Sound Media atom, |
1713 | // except with a version field of 1: |
1714 | int64_t initFilePosn = TellFile64(fOutFid); |
1715 | fCurrentIOState->fQTAudioDataType = "Qclp" ; |
1716 | fCurrentIOState->fQTSoundSampleVersion = 1; |
1717 | unsigned size = addAtom_soundMediaGeneral(); |
1718 | |
1719 | // Next, add the four fields that are particular to version 1: |
1720 | // (Later, parameterize these #####) |
1721 | size += addWord(0x000000a0); // samples per packet |
1722 | size += addWord(0x00000000); // ??? |
1723 | size += addWord(0x00000000); // ??? |
1724 | size += addWord(0x00000002); // bytes per sample (uncompressed) |
1725 | |
1726 | // Other special fields are in a 'wave' atom that follows: |
1727 | size += addAtom_wave(); |
1728 | addAtomEnd; |
1729 | |
1730 | addAtom(wave); |
1731 | size += addAtom_frma(); |
1732 | if (strcmp(fCurrentIOState->fQTAudioDataType, "Qclp" ) == 0) { |
1733 | size += addWord(0x00000014); // ??? |
1734 | size += add4ByteString("Qclp" ); // ??? |
1735 | if (fCurrentIOState->fQTBytesPerFrame == 35) { |
1736 | size += addAtom_Fclp(); // full-rate QCELP |
1737 | } else { |
1738 | size += addAtom_Hclp(); // half-rate QCELP |
1739 | } // what about other QCELP 'rates'??? ##### |
1740 | size += addWord(0x00000008); // ??? |
1741 | size += addWord(0x00000000); // ??? |
1742 | size += addWord(0x00000000); // ??? |
1743 | size += addWord(0x00000008); // ??? |
1744 | } else if (strcmp(fCurrentIOState->fQTAudioDataType, "mp4a" ) == 0) { |
1745 | size += addWord(0x0000000c); // ??? |
1746 | size += add4ByteString("mp4a" ); // ??? |
1747 | size += addWord(0x00000000); // ??? |
1748 | size += addAtom_esds(); // ESDescriptor |
1749 | size += addWord(0x00000008); // ??? |
1750 | size += addWord(0x00000000); // ??? |
1751 | } |
1752 | addAtomEnd; |
1753 | |
1754 | addAtom(frma); |
1755 | size += add4ByteString(fCurrentIOState->fQTAudioDataType); // ??? |
1756 | addAtomEnd; |
1757 | |
1758 | addAtom(Fclp); |
1759 | size += addWord(0x00000000); // ??? |
1760 | addAtomEnd; |
1761 | |
1762 | addAtom(Hclp); |
1763 | size += addWord(0x00000000); // ??? |
1764 | addAtomEnd; |
1765 | |
1766 | unsigned QuickTimeFileSink::addAtom_mp4a() { |
1767 | unsigned size = 0; |
1768 | // The beginning of this atom looks just like a general Sound Media atom, |
1769 | // except with a version field of 1: |
1770 | int64_t initFilePosn = TellFile64(fOutFid); |
1771 | fCurrentIOState->fQTAudioDataType = "mp4a" ; |
1772 | |
1773 | if (fGenerateMP4Format) { |
1774 | fCurrentIOState->fQTSoundSampleVersion = 0; |
1775 | size = addAtom_soundMediaGeneral(); |
1776 | size += addAtom_esds(); |
1777 | } else { |
1778 | fCurrentIOState->fQTSoundSampleVersion = 1; |
1779 | size = addAtom_soundMediaGeneral(); |
1780 | |
1781 | // Next, add the four fields that are particular to version 1: |
1782 | // (Later, parameterize these #####) |
1783 | size += addWord(fCurrentIOState->fQTTimeUnitsPerSample); |
1784 | size += addWord(0x00000001); // ??? |
1785 | size += addWord(0x00000001); // ??? |
1786 | size += addWord(0x00000002); // bytes per sample (uncompressed) |
1787 | |
1788 | // Other special fields are in a 'wave' atom that follows: |
1789 | size += addAtom_wave(); |
1790 | } |
1791 | addAtomEnd; |
1792 | |
1793 | addAtom(esds); |
1794 | //##### |
1795 | MediaSubsession& subsession = fCurrentIOState->fOurSubsession; |
1796 | if (strcmp(subsession.mediumName(), "audio" ) == 0) { |
1797 | // MPEG-4 audio |
1798 | size += addWord(0x00000000); // ??? |
1799 | size += addWord(0x03808080); // ??? |
1800 | size += addWord(0x2a000000); // ??? |
1801 | size += addWord(0x04808080); // ??? |
1802 | size += addWord(0x1c401500); // ??? |
1803 | size += addWord(0x18000000); // ??? |
1804 | size += addWord(0x6d600000); // ??? |
1805 | size += addWord(0x6d600580); // ??? |
1806 | size += addByte(0x80); size += addByte(0x80); // ??? |
1807 | } else if (strcmp(subsession.mediumName(), "video" ) == 0) { |
1808 | // MPEG-4 video |
1809 | size += addWord(0x00000000); // ??? |
1810 | size += addWord(0x03330000); // ??? |
1811 | size += addWord(0x1f042b20); // ??? |
1812 | size += addWord(0x1104fd46); // ??? |
1813 | size += addWord(0x000d4e10); // ??? |
1814 | size += addWord(0x000d4e10); // ??? |
1815 | size += addByte(0x05); // ??? |
1816 | } |
1817 | |
1818 | // Add the source's 'config' information: |
1819 | unsigned configSize; |
1820 | unsigned char* config |
1821 | = parseGeneralConfigStr(subsession.fmtp_config(), configSize); |
1822 | size += addByte(configSize); |
1823 | for (unsigned i = 0; i < configSize; ++i) { |
1824 | size += addByte(config[i]); |
1825 | } |
1826 | delete[] config; |
1827 | |
1828 | if (strcmp(subsession.mediumName(), "audio" ) == 0) { |
1829 | // MPEG-4 audio |
1830 | size += addWord(0x06808080); // ??? |
1831 | size += addHalfWord(0x0102); // ??? |
1832 | } else { |
1833 | // MPEG-4 video |
1834 | size += addHalfWord(0x0601); // ??? |
1835 | size += addByte(0x02); // ??? |
1836 | } |
1837 | //##### |
1838 | addAtomEnd; |
1839 | |
1840 | addAtom(srcq); |
1841 | //##### |
1842 | size += addWord(0x00000040); // ??? |
1843 | //##### |
1844 | addAtomEnd; |
1845 | |
1846 | addAtom(h263); |
1847 | // General sample description fields: |
1848 | size += addWord(0x00000000); // Reserved |
1849 | size += addWord(0x00000001); // Reserved+Data reference index |
1850 | // Video sample description fields: |
1851 | size += addWord(0x00020001); // Version+Revision level |
1852 | size += add4ByteString("appl" ); // Vendor |
1853 | size += addWord(0x00000000); // Temporal quality |
1854 | size += addWord(0x000002fc); // Spatial quality |
1855 | unsigned const widthAndHeight = (fMovieWidth<<16)|fMovieHeight; |
1856 | size += addWord(widthAndHeight); // Width+height |
1857 | size += addWord(0x00480000); // Horizontal resolution |
1858 | size += addWord(0x00480000); // Vertical resolution |
1859 | size += addWord(0x00000000); // Data size |
1860 | size += addWord(0x00010548); // Frame count+Compressor name (start) |
1861 | // "H.263" |
1862 | size += addWord(0x2e323633); // Compressor name (continued) |
1863 | size += addZeroWords(6); // Compressor name (continued - zero) |
1864 | size += addWord(0x00000018); // Compressor name (final)+Depth |
1865 | size += addHalfWord(0xffff); // Color table id |
1866 | addAtomEnd; |
1867 | |
1868 | addAtom(avc1); |
1869 | // General sample description fields: |
1870 | size += addWord(0x00000000); // Reserved |
1871 | size += addWord(0x00000001); // Reserved+Data reference index |
1872 | // Video sample description fields: |
1873 | size += addWord(0x00000000); // Version+Revision level |
1874 | size += add4ByteString("appl" ); // Vendor |
1875 | size += addWord(0x00000000); // Temporal quality |
1876 | size += addWord(0x00000000); // Spatial quality |
1877 | unsigned const widthAndHeight = (fMovieWidth<<16)|fMovieHeight; |
1878 | size += addWord(widthAndHeight); // Width+height |
1879 | size += addWord(0x00480000); // Horizontal resolution |
1880 | size += addWord(0x00480000); // Vertical resolution |
1881 | size += addWord(0x00000000); // Data size |
1882 | size += addWord(0x00010548); // Frame count+Compressor name (start) |
1883 | // "H.264" |
1884 | size += addWord(0x2e323634); // Compressor name (continued) |
1885 | size += addZeroWords(6); // Compressor name (continued - zero) |
1886 | size += addWord(0x00000018); // Compressor name (final)+Depth |
1887 | size += addHalfWord(0xffff); // Color table id |
1888 | size += addAtom_avcC(); |
1889 | addAtomEnd; |
1890 | |
1891 | addAtom(avcC); |
1892 | // Begin by Base-64 decoding the "sprop" parameter sets strings: |
1893 | char* psets = strDup(fCurrentIOState->fOurSubsession.fmtp_spropparametersets()); |
1894 | if (psets == NULL) return 0; |
1895 | |
1896 | size_t comma_pos = strcspn(psets, "," ); |
1897 | psets[comma_pos] = '\0'; |
1898 | char const* sps_b64 = psets; |
1899 | char const* pps_b64 = &psets[comma_pos+1]; |
1900 | unsigned sps_count; |
1901 | unsigned char* sps_data = base64Decode(sps_b64, sps_count, false); |
1902 | unsigned pps_count; |
1903 | unsigned char* pps_data = base64Decode(pps_b64, pps_count, false); |
1904 | |
1905 | // Then add the decoded data: |
1906 | size += addByte(0x01); // configuration version |
1907 | size += addByte(sps_data[1]); // profile |
1908 | size += addByte(sps_data[2]); // profile compat |
1909 | size += addByte(sps_data[3]); // level |
1910 | size += addByte(0xff); /* 0b11111100 | lengthsize = 0x11 */ |
1911 | size += addByte(0xe0 | (sps_count > 0 ? 1 : 0) ); |
1912 | if (sps_count > 0) { |
1913 | size += addHalfWord(sps_count); |
1914 | for (unsigned i = 0; i < sps_count; i++) { |
1915 | size += addByte(sps_data[i]); |
1916 | } |
1917 | } |
1918 | size += addByte(pps_count > 0 ? 1 : 0); |
1919 | if (pps_count > 0) { |
1920 | size += addHalfWord(pps_count); |
1921 | for (unsigned i = 0; i < pps_count; i++) { |
1922 | size += addByte(pps_data[i]); |
1923 | } |
1924 | } |
1925 | |
1926 | // Finally, delete the data that we allocated: |
1927 | delete[] pps_data; delete[] sps_data; |
1928 | delete[] psets; |
1929 | addAtomEnd; |
1930 | |
1931 | addAtom(mp4v); |
1932 | // General sample description fields: |
1933 | size += addWord(0x00000000); // Reserved |
1934 | size += addWord(0x00000001); // Reserved+Data reference index |
1935 | // Video sample description fields: |
1936 | size += addWord(0x00020001); // Version+Revision level |
1937 | size += add4ByteString("appl" ); // Vendor |
1938 | size += addWord(0x00000200); // Temporal quality |
1939 | size += addWord(0x00000400); // Spatial quality |
1940 | unsigned const widthAndHeight = (fMovieWidth<<16)|fMovieHeight; |
1941 | size += addWord(widthAndHeight); // Width+height |
1942 | size += addWord(0x00480000); // Horizontal resolution |
1943 | size += addWord(0x00480000); // Vertical resolution |
1944 | size += addWord(0x00000000); // Data size |
1945 | size += addWord(0x00010c4d); // Frame count+Compressor name (start) |
1946 | // "MPEG-4 Video" |
1947 | size += addWord(0x5045472d); // Compressor name (continued) |
1948 | size += addWord(0x34205669); // Compressor name (continued) |
1949 | size += addWord(0x64656f00); // Compressor name (continued) |
1950 | size += addZeroWords(4); // Compressor name (continued - zero) |
1951 | size += addWord(0x00000018); // Compressor name (final)+Depth |
1952 | size += addHalfWord(0xffff); // Color table id |
1953 | size += addAtom_esds(); // ESDescriptor |
1954 | size += addWord(0x00000000); // ??? |
1955 | addAtomEnd; |
1956 | |
1957 | unsigned QuickTimeFileSink::addAtom_rtp() { |
1958 | int64_t initFilePosn = TellFile64(fOutFid); |
1959 | unsigned size = addAtomHeader("rtp " ); |
1960 | |
1961 | size += addWord(0x00000000); // Reserved (1st 4 bytes) |
1962 | size += addWord(0x00000001); // Reserved (last 2 bytes) + Data ref index |
1963 | size += addWord(0x00010001); // Hint track version + Last compat htv |
1964 | size += addWord(1450); // Max packet size |
1965 | |
1966 | size += addAtom_tims(); |
1967 | addAtomEnd; |
1968 | |
1969 | addAtom(tims); |
1970 | size += addWord(fCurrentIOState->fOurSubsession.rtpTimestampFrequency()); |
1971 | addAtomEnd; |
1972 | |
1973 | addAtom(stts); // Time-to-Sample |
1974 | size += addWord(0x00000000); // Version+flags |
1975 | |
1976 | // First, add a dummy "Number of entries" field |
1977 | // (and remember its position). We'll fill this field in later: |
1978 | int64_t numEntriesPosition = TellFile64(fOutFid); |
1979 | size += addWord(0); // dummy for "Number of entries" |
1980 | |
1981 | // Then, run through the chunk descriptors, and enter the entries |
1982 | // in this (compressed) Time-to-Sample table: |
1983 | unsigned numEntries = 0, numSamplesSoFar = 0; |
1984 | unsigned prevSampleDuration = 0; |
1985 | unsigned const samplesPerFrame = fCurrentIOState->fQTSamplesPerFrame; |
1986 | ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk; |
1987 | while (chunk != NULL) { |
1988 | unsigned const sampleDuration = chunk->fFrameDuration/samplesPerFrame; |
1989 | if (sampleDuration != prevSampleDuration) { |
1990 | // This chunk will start a new table entry, |
1991 | // so write out the old one (if any): |
1992 | if (chunk != fCurrentIOState->fHeadChunk) { |
1993 | ++numEntries; |
1994 | size += addWord(numSamplesSoFar); // Sample count |
1995 | size += addWord(prevSampleDuration); // Sample duration |
1996 | numSamplesSoFar = 0; |
1997 | } |
1998 | } |
1999 | |
2000 | unsigned const numSamples = chunk->fNumFrames*samplesPerFrame; |
2001 | numSamplesSoFar += numSamples; |
2002 | prevSampleDuration = sampleDuration; |
2003 | chunk = chunk->fNextChunk; |
2004 | } |
2005 | |
2006 | // Then, write out the last entry: |
2007 | ++numEntries; |
2008 | size += addWord(numSamplesSoFar); // Sample count |
2009 | size += addWord(prevSampleDuration); // Sample duration |
2010 | |
2011 | // Now go back and fill in the "Number of entries" field: |
2012 | setWord(numEntriesPosition, numEntries); |
2013 | addAtomEnd; |
2014 | |
2015 | addAtom(stss); // Sync-Sample |
2016 | size += addWord(0x00000000); // Version+flags |
2017 | |
2018 | // First, add a dummy "Number of entries" field |
2019 | // (and remember its position). We'll fill this field in later: |
2020 | int64_t numEntriesPosition = TellFile64(fOutFid); |
2021 | size += addWord(0); // dummy for "Number of entries" |
2022 | |
2023 | unsigned numEntries = 0, numSamplesSoFar = 0; |
2024 | if (fCurrentIOState->fHeadSyncFrame != NULL) { |
2025 | SyncFrame* currentSyncFrame = fCurrentIOState->fHeadSyncFrame; |
2026 | |
2027 | // First, count the number of frames (to use as a sanity check; see below): |
2028 | unsigned totNumFrames = 0; |
2029 | for (ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk; chunk != NULL; chunk = chunk->fNextChunk) totNumFrames += chunk->fNumFrames; |
2030 | |
2031 | while (currentSyncFrame != NULL) { |
2032 | if (currentSyncFrame->sfFrameNum >= totNumFrames) break; // sanity check |
2033 | |
2034 | ++numEntries; |
2035 | size += addWord(currentSyncFrame->sfFrameNum); |
2036 | currentSyncFrame = currentSyncFrame->nextSyncFrame; |
2037 | } |
2038 | } else { |
2039 | // First, run through the chunk descriptors, counting up the total number of samples: |
2040 | unsigned const samplesPerFrame = fCurrentIOState->fQTSamplesPerFrame; |
2041 | ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk; |
2042 | while (chunk != NULL) { |
2043 | unsigned const numSamples = chunk->fNumFrames*samplesPerFrame; |
2044 | numSamplesSoFar += numSamples; |
2045 | chunk = chunk->fNextChunk; |
2046 | } |
2047 | |
2048 | // Then, write out the sample numbers that we deem correspond to 'sync samples': |
2049 | unsigned i; |
2050 | for (i = 0; i < numSamplesSoFar; i += 12) { |
2051 | // For an explanation of the constant "12", see http://lists.live555.com/pipermail/live-devel/2009-July/010969.html |
2052 | // (Perhaps we should really try to keep track of which 'samples' ('frames' for video) really are 'key frames'?) |
2053 | size += addWord(i+1); |
2054 | ++numEntries; |
2055 | } |
2056 | |
2057 | // Then, write out the last entry (if we haven't already done so): |
2058 | if (i != (numSamplesSoFar - 1)) { |
2059 | size += addWord(numSamplesSoFar); |
2060 | ++numEntries; |
2061 | } |
2062 | } |
2063 | |
2064 | // Now go back and fill in the "Number of entries" field: |
2065 | setWord(numEntriesPosition, numEntries); |
2066 | addAtomEnd; |
2067 | |
2068 | addAtom(stsc); // Sample-to-Chunk |
2069 | size += addWord(0x00000000); // Version+flags |
2070 | |
2071 | // First, add a dummy "Number of entries" field |
2072 | // (and remember its position). We'll fill this field in later: |
2073 | int64_t numEntriesPosition = TellFile64(fOutFid); |
2074 | size += addWord(0); // dummy for "Number of entries" |
2075 | |
2076 | // Then, run through the chunk descriptors, and enter the entries |
2077 | // in this (compressed) Sample-to-Chunk table: |
2078 | unsigned numEntries = 0, chunkNumber = 0; |
2079 | unsigned prevSamplesPerChunk = ~0; |
2080 | unsigned const samplesPerFrame = fCurrentIOState->fQTSamplesPerFrame; |
2081 | ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk; |
2082 | while (chunk != NULL) { |
2083 | ++chunkNumber; |
2084 | unsigned const samplesPerChunk = chunk->fNumFrames*samplesPerFrame; |
2085 | if (samplesPerChunk != prevSamplesPerChunk) { |
2086 | // This chunk will be a new table entry: |
2087 | ++numEntries; |
2088 | size += addWord(chunkNumber); // Chunk number |
2089 | size += addWord(samplesPerChunk); // Samples per chunk |
2090 | size += addWord(0x00000001); // Sample description ID |
2091 | |
2092 | prevSamplesPerChunk = samplesPerChunk; |
2093 | } |
2094 | chunk = chunk->fNextChunk; |
2095 | } |
2096 | |
2097 | // Now go back and fill in the "Number of entries" field: |
2098 | setWord(numEntriesPosition, numEntries); |
2099 | addAtomEnd; |
2100 | |
2101 | addAtom(stsz); // Sample Size |
2102 | size += addWord(0x00000000); // Version+flags |
2103 | |
2104 | // Begin by checking whether our chunks all have the same |
2105 | // 'bytes-per-sample'. This determines whether this atom's table |
2106 | // has just a single entry, or multiple entries. |
2107 | Boolean haveSingleEntryTable = True; |
2108 | double firstBPS = 0.0; |
2109 | ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk; |
2110 | while (chunk != NULL) { |
2111 | double bps |
2112 | = (double)(chunk->fFrameSize)/(fCurrentIOState->fQTSamplesPerFrame); |
2113 | if (bps < 1.0) { |
2114 | // I don't think a multiple-entry table would make sense in |
2115 | // this case, so assume a single entry table ??? ##### |
2116 | break; |
2117 | } |
2118 | |
2119 | if (firstBPS == 0.0) { |
2120 | firstBPS = bps; |
2121 | } else if (bps != firstBPS) { |
2122 | haveSingleEntryTable = False; |
2123 | break; |
2124 | } |
2125 | |
2126 | chunk = chunk->fNextChunk; |
2127 | } |
2128 | |
2129 | unsigned sampleSize; |
2130 | if (haveSingleEntryTable) { |
2131 | if (fCurrentIOState->isHintTrack() |
2132 | && fCurrentIOState->fHeadChunk != NULL) { |
2133 | sampleSize = fCurrentIOState->fHeadChunk->fFrameSize |
2134 | / fCurrentIOState->fQTSamplesPerFrame; |
2135 | } else { |
2136 | // The following doesn't seem right, but seems to do the right thing: |
2137 | sampleSize = fCurrentIOState->fQTTimeUnitsPerSample; //??? |
2138 | } |
2139 | } else { |
2140 | sampleSize = 0; // indicates a multiple-entry table |
2141 | } |
2142 | size += addWord(sampleSize); // Sample size |
2143 | unsigned const totNumSamples = fCurrentIOState->fQTTotNumSamples; |
2144 | size += addWord(totNumSamples); // Number of entries |
2145 | |
2146 | if (!haveSingleEntryTable) { |
2147 | // Multiple-entry table: |
2148 | // Run through the chunk descriptors, entering the sample sizes: |
2149 | ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk; |
2150 | while (chunk != NULL) { |
2151 | unsigned numSamples |
2152 | = chunk->fNumFrames*(fCurrentIOState->fQTSamplesPerFrame); |
2153 | unsigned sampleSize |
2154 | = chunk->fFrameSize/(fCurrentIOState->fQTSamplesPerFrame); |
2155 | for (unsigned i = 0; i < numSamples; ++i) { |
2156 | size += addWord(sampleSize); |
2157 | } |
2158 | |
2159 | chunk = chunk->fNextChunk; |
2160 | } |
2161 | } |
2162 | addAtomEnd; |
2163 | |
2164 | addAtom(co64); // Chunk Offset |
2165 | size += addWord(0x00000000); // Version+flags |
2166 | size += addWord(fCurrentIOState->fNumChunks); // Number of entries |
2167 | |
2168 | // Run through the chunk descriptors, entering the file offsets: |
2169 | ChunkDescriptor* chunk = fCurrentIOState->fHeadChunk; |
2170 | while (chunk != NULL) { |
2171 | size += addWord64(chunk->fOffsetInFile); |
2172 | |
2173 | chunk = chunk->fNextChunk; |
2174 | } |
2175 | addAtomEnd; |
2176 | |
2177 | addAtom(udta); |
2178 | size += addAtom_name(); |
2179 | size += addAtom_hnti(); |
2180 | size += addAtom_hinf(); |
2181 | addAtomEnd; |
2182 | |
2183 | addAtom(name); |
2184 | char description[100]; |
2185 | sprintf(description, "Hinted %s track" , |
2186 | fCurrentIOState->fOurSubsession.mediumName()); |
2187 | size += addArbitraryString(description, False); // name of object |
2188 | addAtomEnd; |
2189 | |
2190 | addAtom(hnti); |
2191 | size += addAtom_sdp(); |
2192 | addAtomEnd; |
2193 | |
2194 | unsigned QuickTimeFileSink::addAtom_sdp() { |
2195 | int64_t initFilePosn = TellFile64(fOutFid); |
2196 | unsigned size = addAtomHeader("sdp " ); |
2197 | |
2198 | // Add this subsession's SDP lines: |
2199 | char const* sdpLines = fCurrentIOState->fOurSubsession.savedSDPLines(); |
2200 | // We need to change any "a=control:trackID=" values to be this |
2201 | // track's actual track id: |
2202 | char* newSDPLines = new char[strlen(sdpLines)+100/*overkill*/]; |
2203 | char const* searchStr = "a=control:trackid=" ; |
2204 | Boolean foundSearchString = False; |
2205 | char const *p1, *p2, *p3; |
2206 | for (p1 = sdpLines; *p1 != '\0'; ++p1) { |
2207 | for (p2 = p1,p3 = searchStr; tolower(*p2) == *p3; ++p2,++p3) {} |
2208 | if (*p3 == '\0') { |
2209 | // We found the end of the search string, at p2. |
2210 | int beforeTrackNumPosn = p2-sdpLines; |
2211 | // Look for the subsequent track number, and skip over it: |
2212 | int trackNumLength; |
2213 | if (sscanf(p2, " %*d%n" , &trackNumLength) < 0) break; |
2214 | int afterTrackNumPosn = beforeTrackNumPosn + trackNumLength; |
2215 | |
2216 | // Replace the old track number with the correct one: |
2217 | int i; |
2218 | for (i = 0; i < beforeTrackNumPosn; ++i) newSDPLines[i] = sdpLines[i]; |
2219 | sprintf(&newSDPLines[i], "%d" , fCurrentIOState->fTrackID); |
2220 | i = afterTrackNumPosn; |
2221 | int j = i + strlen(&newSDPLines[i]); |
2222 | while (1) { |
2223 | if ((newSDPLines[j] = sdpLines[i]) == '\0') break; |
2224 | ++i; ++j; |
2225 | } |
2226 | |
2227 | foundSearchString = True; |
2228 | break; |
2229 | } |
2230 | } |
2231 | |
2232 | if (!foundSearchString) { |
2233 | // Because we didn't find a "a=control:trackID=<trackId>" line, |
2234 | // add one of our own: |
2235 | sprintf(newSDPLines, "%s%s%d\r\n" , |
2236 | sdpLines, searchStr, fCurrentIOState->fTrackID); |
2237 | } |
2238 | |
2239 | size += addArbitraryString(newSDPLines, False); |
2240 | delete[] newSDPLines; |
2241 | addAtomEnd; |
2242 | |
2243 | addAtom(hinf); |
2244 | size += addAtom_totl(); |
2245 | size += addAtom_npck(); |
2246 | size += addAtom_tpay(); |
2247 | size += addAtom_trpy(); |
2248 | size += addAtom_nump(); |
2249 | size += addAtom_tpyl(); |
2250 | // Is 'maxr' required? ##### |
2251 | size += addAtom_dmed(); |
2252 | size += addAtom_dimm(); |
2253 | size += addAtom_drep(); |
2254 | size += addAtom_tmin(); |
2255 | size += addAtom_tmax(); |
2256 | size += addAtom_pmax(); |
2257 | size += addAtom_dmax(); |
2258 | size += addAtom_payt(); |
2259 | addAtomEnd; |
2260 | |
2261 | addAtom(totl); |
2262 | size += addWord(fCurrentIOState->fHINF.trpy.lo); |
2263 | addAtomEnd; |
2264 | |
2265 | addAtom(npck); |
2266 | size += addWord(fCurrentIOState->fHINF.nump.lo); |
2267 | addAtomEnd; |
2268 | |
2269 | addAtom(tpay); |
2270 | size += addWord(fCurrentIOState->fHINF.tpyl.lo); |
2271 | addAtomEnd; |
2272 | |
2273 | addAtom(trpy); |
2274 | size += addWord(fCurrentIOState->fHINF.trpy.hi); |
2275 | size += addWord(fCurrentIOState->fHINF.trpy.lo); |
2276 | addAtomEnd; |
2277 | |
2278 | addAtom(nump); |
2279 | size += addWord(fCurrentIOState->fHINF.nump.hi); |
2280 | size += addWord(fCurrentIOState->fHINF.nump.lo); |
2281 | addAtomEnd; |
2282 | |
2283 | addAtom(tpyl); |
2284 | size += addWord(fCurrentIOState->fHINF.tpyl.hi); |
2285 | size += addWord(fCurrentIOState->fHINF.tpyl.lo); |
2286 | addAtomEnd; |
2287 | |
2288 | addAtom(dmed); |
2289 | size += addWord(fCurrentIOState->fHINF.dmed.hi); |
2290 | size += addWord(fCurrentIOState->fHINF.dmed.lo); |
2291 | addAtomEnd; |
2292 | |
2293 | addAtom(dimm); |
2294 | size += addWord(fCurrentIOState->fHINF.dimm.hi); |
2295 | size += addWord(fCurrentIOState->fHINF.dimm.lo); |
2296 | addAtomEnd; |
2297 | |
2298 | addAtom(drep); |
2299 | size += addWord(0); |
2300 | size += addWord(0); |
2301 | addAtomEnd; |
2302 | |
2303 | addAtom(tmin); |
2304 | size += addWord(0); |
2305 | addAtomEnd; |
2306 | |
2307 | addAtom(tmax); |
2308 | size += addWord(0); |
2309 | addAtomEnd; |
2310 | |
2311 | addAtom(pmax); |
2312 | size += addWord(fCurrentIOState->fHINF.pmax); |
2313 | addAtomEnd; |
2314 | |
2315 | addAtom(dmax); |
2316 | size += addWord(fCurrentIOState->fHINF.dmax); |
2317 | addAtomEnd; |
2318 | |
2319 | addAtom(payt); |
2320 | MediaSubsession& ourSubsession = fCurrentIOState->fOurSubsession; |
2321 | RTPSource* rtpSource = ourSubsession.rtpSource(); |
2322 | if (rtpSource != NULL) { |
2323 | size += addWord(rtpSource->rtpPayloadFormat()); |
2324 | |
2325 | // Also, add a 'rtpmap' string: <mime-subtype>/<rtp-frequency> |
2326 | unsigned rtpmapStringLength = strlen(ourSubsession.codecName()) + 20; |
2327 | char* rtpmapString = new char[rtpmapStringLength]; |
2328 | sprintf(rtpmapString, "%s/%d" , |
2329 | ourSubsession.codecName(), rtpSource->timestampFrequency()); |
2330 | size += addArbitraryString(rtpmapString); |
2331 | delete[] rtpmapString; |
2332 | } |
2333 | addAtomEnd; |
2334 | |
2335 | // A dummy atom (with name "????"): |
2336 | unsigned QuickTimeFileSink::addAtom_dummy() { |
2337 | int64_t initFilePosn = TellFile64(fOutFid); |
2338 | unsigned size = addAtomHeader("????" ); |
2339 | addAtomEnd; |
2340 | |