EbmlWriter.cpp@ 68451

Last change on this file since 68451 was 68451, checked in by vboxsync, 7 years ago

VideoRec: Main/EbmlWriter: More stuff for audio support.

Property svn:eol-style set to native
Property svn:keywords set to Author Date Id Revision

Property svn:mergeinfo set to (toggle deleted branches)

/branches/VBox-3.0/src/VBox/Frontends/VBoxHeadless/VideoCapture/EbmlWriter.cpp	58652,70973
/branches/VBox-3.2/src/VBox/Frontends/VBoxHeadless/VideoCapture/EbmlWriter.cpp	66309,66318
/branches/VBox-4.0/src/VBox/Frontends/VBoxHeadless/VideoCapture/EbmlWriter.cpp	70873
/branches/VBox-4.1/src/VBox/Frontends/VBoxHeadless/VideoCapture/EbmlWriter.cpp	74233
/branches/VBox-4.2/src/VBox/Main/src-client/EbmlWriter.cpp	91503-91504,91506-91508,91510,91514-91515,91521
/branches/VBox-4.3/src/VBox/Main/src-client/EbmlWriter.cpp	91223
/branches/VBox-4.3/trunk/src/VBox/Main/src-client/EbmlWriter.cpp	91223
/branches/dsen/gui/src/VBox/Frontends/VBoxHeadless/VideoCapture/EbmlWriter.cpp	79076-79078,79089,79109-79110,79112-79113,79127-79130,79134,79141,79151,79155,79157-79159,79193,79197
/branches/dsen/gui2/src/VBox/Frontends/VBoxHeadless/VideoCapture/EbmlWriter.cpp	79224,79228,79233,79235,79258,79262-79263,79273,79341,79345,79354,79357,79387-79388,79559-79569,79572-79573,79578,79581-79582,79590-79591,79598-79599,79602-79603,79605-79606,79632,79635,79637,79644
/branches/dsen/gui3/src/VBox/Frontends/VBoxHeadless/VideoCapture/EbmlWriter.cpp	79645-79692

File size: 38.0 KB

Line
1	/* $Id: EbmlWriter.cpp 68451 2017-08-17 19:54:52Z vboxsync $ */
2	/** @file
3	* EbmlWriter.cpp - EBML writer + WebM container handling.
4	*/
5
6	/*
7	* Copyright (C) 2013-2017 Oracle Corporation
8	*
9	* This file is part of VirtualBox Open Source Edition (OSE), as
10	* available from http://www.virtualbox.org. This file is free software;
11	* you can redistribute it and/or modify it under the terms of the GNU
12	* General Public License (GPL) as published by the Free Software
13	* Foundation, in version 2 as it comes in the "COPYING" file of the
14	* VirtualBox OSE distribution. VirtualBox OSE is distributed in the
15	* hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
16	*/
17
18	/**
19	* For more information, see:
20	* - https://w3c.github.io/media-source/webm-byte-stream-format.html
21	* - https://www.webmproject.org/docs/container/#muxer-guidelines
22	*/
23
24	#define LOG_GROUP LOG_GROUP_MAIN_DISPLAY
25	#include "LoggingNew.h"
26
27	#include <list>
28	#include <map>
29	#include <stack>
30
31	#include <math.h> /* For lround.h. */
32
33	#include <iprt/asm.h>
34	#include <iprt/buildconfig.h>
35	#include <iprt/cdefs.h>
36	#include <iprt/err.h>
37	#include <iprt/file.h>
38	#include <iprt/rand.h>
39	#include <iprt/string.h>
40
41	#include <VBox/log.h>
42	#include <VBox/version.h>
43
44	#include "EbmlWriter.h"
45	#include "EbmlMkvIDs.h"
46
47
48	class Ebml
49	{
50	public:
51	typedef uint32_t EbmlClassId;
52
53	private:
54
55	struct EbmlSubElement
56	{
57	uint64_t offset;
58	EbmlClassId classId;
59	EbmlSubElement(uint64_t offs, EbmlClassId cid) : offset(offs), classId(cid) {}
60	};
61
62	/** Stack of EBML sub elements. */
63	std::stack<EbmlSubElement> m_Elements;
64	/** The file's handle. */
65	RTFILE m_hFile;
66	/** The file's name (path). */
67	Utf8Str m_strFile;
68
69	public:
70
71	Ebml(void)
72	: m_hFile(NIL_RTFILE) { }
73
74	virtual ~Ebml(void) { close(); }
75
76	public:
77
78	/** Creates EBML output file. */
79	inline int create(const char *a_pszFilename, uint64_t fOpen)
80	{
81	int rc = RTFileOpen(&m_hFile, a_pszFilename, fOpen);
82	if (RT_SUCCESS(rc))
83	{
84	m_strFile = a_pszFilename;
85	}
86
87	return rc;
88	}
89
90	/** Returns the file name. */
91	inline const Utf8Str& getFileName(void)
92	{
93	return m_strFile;
94	}
95
96	/** Returns file size. */
97	inline uint64_t getFileSize(void)
98	{
99	return RTFileTell(m_hFile);
100	}
101
102	/** Get reference to file descriptor */
103	inline const RTFILE &getFile(void)
104	{
105	return m_hFile;
106	}
107
108	/** Returns available space on storage. */
109	inline uint64_t getAvailableSpace(void)
110	{
111	RTFOFF pcbFree;
112	int rc = RTFileQueryFsSizes(m_hFile, NULL, &pcbFree, 0, 0);
113	return (RT_SUCCESS(rc)? (uint64_t)pcbFree : UINT64_MAX);
114	}
115
116	/** Closes the file. */
117	inline void close(void)
118	{
119	if (!isOpen())
120	return;
121
122	AssertMsg(m_Elements.size() == 0,
123	("%zu elements are not closed yet (next element to close is 0x%x)\n",
124	m_Elements.size(), m_Elements.top().classId));
125
126	RTFileClose(m_hFile);
127	m_hFile = NIL_RTFILE;
128
129	m_strFile = "";
130	}
131
132	/**
133	* Returns whether the file is open or not.
134	*
135	* @returns True if open, false if not.
136	*/
137	inline bool isOpen(void)
138	{
139	return RTFileIsValid(m_hFile);
140	}
141
142	/** Starts an EBML sub-element. */
143	inline Ebml &subStart(EbmlClassId classId)
144	{
145	writeClassId(classId);
146	/* store the current file offset. */
147	m_Elements.push(EbmlSubElement(RTFileTell(m_hFile), classId));
148	/* Indicates that size of the element
149	* is unkown (as according to EBML specs).
150	*/
151	writeUnsignedInteger(UINT64_C(0x01FFFFFFFFFFFFFF));
152	return *this;
153	}
154
155	/** Ends an EBML sub-element. */
156	inline Ebml &subEnd(EbmlClassId classId)
157	{
158	#ifdef VBOX_STRICT
159	/* Class ID on the top of the stack should match the class ID passed
160	* to the function. Otherwise it may mean that we have a bug in the code.
161	*/
162	AssertMsg(!m_Elements.empty(), ("No elements to close anymore\n"));
163	AssertMsg(m_Elements.top().classId == classId,
164	("Ending sub element 0x%x is in wrong order (next to close is 0x%x)\n", classId, m_Elements.top().classId));
165	#else
166	RT_NOREF(classId);
167	#endif
168
169	uint64_t uPos = RTFileTell(m_hFile);
170	uint64_t uSize = uPos - m_Elements.top().offset - 8;
171	RTFileSeek(m_hFile, m_Elements.top().offset, RTFILE_SEEK_BEGIN, NULL);
172
173	/* Make sure that size will be serialized as uint64_t. */
174	writeUnsignedInteger(uSize \| UINT64_C(0x0100000000000000));
175	RTFileSeek(m_hFile, uPos, RTFILE_SEEK_BEGIN, NULL);
176	m_Elements.pop();
177	return *this;
178	}
179
180	/** Serializes a null-terminated string. */
181	inline Ebml &serializeString(EbmlClassId classId, const char *str)
182	{
183	writeClassId(classId);
184	uint64_t size = strlen(str);
185	writeSize(size);
186	write(str, size);
187	return *this;
188	}
189
190	/* Serializes an UNSIGNED integer
191	* If size is zero then it will be detected automatically. */
192	inline Ebml &serializeUnsignedInteger(EbmlClassId classId, uint64_t parm, size_t size = 0)
193	{
194	writeClassId(classId);
195	if (!size) size = getSizeOfUInt(parm);
196	writeSize(size);
197	writeUnsignedInteger(parm, size);
198	return *this;
199	}
200
201	/** Serializes a floating point value.
202	*
203	* Only 8-bytes double precision values are supported
204	* by this function.
205	*/
206	inline Ebml &serializeFloat(EbmlClassId classId, float value)
207	{
208	writeClassId(classId);
209	Assert(sizeof(uint32_t) == sizeof(float));
210	writeSize(sizeof(float));
211
212	union
213	{
214	float f;
215	uint8_t u8[4];
216	} u;
217
218	u.f = value;
219
220	for (int i = 3; i >= 0; i--) /* Converts values to big endian. */
221	write(&u.u8[i], 1);
222
223	return *this;
224	}
225
226	/** Serializes binary data. */
227	inline Ebml &serializeData(EbmlClassId classId, const void *pvData, size_t cbData)
228	{
229	writeClassId(classId);
230	writeSize(cbData);
231	write(pvData, cbData);
232	return *this;
233	}
234
235	/** Writes raw data to file. */
236	inline int write(const void *data, size_t size)
237	{
238	return RTFileWrite(m_hFile, data, size, NULL);
239	}
240
241	/** Writes an unsigned integer of variable of fixed size. */
242	inline void writeUnsignedInteger(uint64_t value, size_t size = sizeof(uint64_t))
243	{
244	/* convert to big-endian */
245	value = RT_H2BE_U64(value);
246	write(reinterpret_cast<uint8_t*>(&value) + sizeof(value) - size, size);
247	}
248
249	/** Writes EBML class ID to file.
250	*
251	* EBML ID already has a UTF8-like represenation
252	* so getSizeOfUInt is used to determine
253	* the number of its bytes.
254	*/
255	inline void writeClassId(EbmlClassId parm)
256	{
257	writeUnsignedInteger(parm, getSizeOfUInt(parm));
258	}
259
260	/** Writes data size value. */
261	inline void writeSize(uint64_t parm)
262	{
263	/* The following expression defines the size of the value that will be serialized
264	* as an EBML UTF-8 like integer (with trailing bits represeting its size):
265	1xxx xxxx - value 0 to 2^7-2
266	01xx xxxx xxxx xxxx - value 0 to 2^14-2
267	001x xxxx xxxx xxxx xxxx xxxx - value 0 to 2^21-2
268	0001 xxxx xxxx xxxx xxxx xxxx xxxx xxxx - value 0 to 2^28-2
269	0000 1xxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx - value 0 to 2^35-2
270	0000 01xx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx - value 0 to 2^42-2
271	0000 001x xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx - value 0 to 2^49-2
272	0000 0001 xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx xxxx - value 0 to 2^56-2
273	*/
274	size_t size = 8 - ! (parm & (UINT64_MAX << 49)) - ! (parm & (UINT64_MAX << 42)) -
275	! (parm & (UINT64_MAX << 35)) - ! (parm & (UINT64_MAX << 28)) -
276	! (parm & (UINT64_MAX << 21)) - ! (parm & (UINT64_MAX << 14)) -
277	! (parm & (UINT64_MAX << 7));
278	/* One is subtracted in order to avoid loosing significant bit when size = 8. */
279	uint64_t mask = RT_BIT_64(size * 8 - 1);
280	writeUnsignedInteger((parm & (((mask << 1) - 1) >> size)) \| (mask >> (size - 1)), size);
281	}
282
283	/** Size calculation for variable size UNSIGNED integer.
284	*
285	* The function defines the size of the number by trimming
286	* consequent trailing zero bytes starting from the most significant.
287	* The following statement is always true:
288	* 1 <= getSizeOfUInt(arg) <= 8.
289	*
290	* Every !(arg & (UINT64_MAX << X)) expression gives one
291	* if an only if all the bits from X to 63 are set to zero.
292	*/
293	static inline size_t getSizeOfUInt(uint64_t arg)
294	{
295	return 8 - ! (arg & (UINT64_MAX << 56)) - ! (arg & (UINT64_MAX << 48)) -
296	! (arg & (UINT64_MAX << 40)) - ! (arg & (UINT64_MAX << 32)) -
297	! (arg & (UINT64_MAX << 24)) - ! (arg & (UINT64_MAX << 16)) -
298	! (arg & (UINT64_MAX << 8));
299	}
300
301	private:
302	void operator=(const Ebml &);
303
304	};
305
306	/** No flags specified. */
307	#define VBOX_WEBM_BLOCK_FLAG_NONE 0
308	/** Invisible block which can be skipped. */
309	#define VBOX_WEBM_BLOCK_FLAG_INVISIBLE 0x08
310	/** The block marks a key frame. */
311	#define VBOX_WEBM_BLOCK_FLAG_KEY_FRAME 0x80
312
313	/** The default timecode scale factor for WebM -- all timecodes in the segments are expressed in ms.
314	* This allows every cluster to have blocks with positive values up to 32.767 seconds. */
315	#define VBOX_WEBM_TIMECODESCALE_FACTOR_MS 1000000
316
317	/** Maximum time (in ms) a cluster can store. */
318	#define VBOX_WEBM_CLUSTER_MAX_LEN_MS INT16_MAX
319
320	/** Maximum time a block can store.
321	* With signed 16-bit timecodes and a default timecode scale of 1ms per unit this makes 65536ms. */
322	#define VBOX_WEBM_BLOCK_MAX_LEN_MS UINT16_MAX
323
324	class WebMWriter_Impl
325	{
326	/**
327	* Track type enumeration.
328	*/
329	enum WebMTrackType
330	{
331	/** Unknown / invalid type. */
332	WebMTrackType_Invalid = 0,
333	/** Only writes audio. */
334	WebMTrackType_Audio = 1,
335	/** Only writes video. */
336	WebMTrackType_Video = 2
337	};
338
339	/**
340	* Structure for keeping a WebM track entry.
341	*/
342	struct WebMTrack
343	{
344	WebMTrack(WebMTrackType a_enmType, uint8_t a_uTrack, uint64_t a_offID)
345	: enmType(a_enmType)
346	, uTrack(a_uTrack)
347	, offUUID(a_offID)
348	, cTotalClusters(0)
349	, cTotalBlocks(0)
350	{
351	uUUID = RTRandU32();
352	}
353
354	/** The type of this track. */
355	WebMTrackType enmType;
356	/** Track parameters. */
357	union
358	{
359	struct
360	{
361	/** Sample rate of input data. */
362	uint32_t uHz;
363	/** Duration of the frame in samples (per channel).
364	* Valid frame size are:
365	*
366	* ms Frame size
367	* 2.5 120
368	* 5 240
369	* 10 480
370	* 20 (Default) 960
371	* 40 1920
372	* 60 2880
373	*/
374	uint16_t framesPerBlock;
375	/** How many milliseconds (ms) one written (simple) block represents. */
376	uint16_t msPerBlock;
377	} Audio;
378	};
379	/** This track's track number. Also used as key in track map. */
380	uint8_t uTrack;
381	/** The track's "UUID".
382	* Needed in case this track gets mux'ed with tracks from other files. Not really unique though. */
383	uint32_t uUUID;
384	/** Absolute offset in file of track UUID.
385	* Needed to write the hash sum within the footer. */
386	uint64_t offUUID;
387	/** Total number of clusters. */
388	uint64_t cTotalClusters;
389	/** Total number of blocks. */
390	uint64_t cTotalBlocks;
391	};
392
393	/**
394	* Structure for keeping a cue point.
395	*/
396	struct WebMCuePoint
397	{
398	WebMCuePoint(WebMTrack *a_pTrack, uint32_t a_tcClusterStart, uint64_t a_offClusterStart)
399	: pTrack(a_pTrack)
400	, tcClusterStart(a_tcClusterStart), offClusterStart(a_offClusterStart) {}
401
402	/** Associated track. */
403	WebMTrack *pTrack;
404	/** Start time code of the related cluster. */
405	uint32_t tcClusterStart;
406	/** Start offset of the related cluster. */
407	uint64_t offClusterStart;
408	};
409
410	/**
411	* Structure for keeping a WebM cluster entry.
412	*/
413	struct WebMCluster
414	{
415	WebMCluster(void)
416	: uID(0)
417	, offCluster(0)
418	, fOpen(false)
419	, tcStartMs(0)
420	, tcEndMs(0)
421	, cBlocks(0) { }
422
423	/** This cluster's ID. */
424	uint64_t uID;
425	/** Absolute offset (in bytes) of current cluster.
426	* Needed for seeking info table. */
427	uint64_t offCluster;
428	/** Whether this cluster element is opened currently. */
429	bool fOpen;
430	/** Timecode (in ms) when starting this cluster. */
431	uint64_t tcStartMs;
432	/** Timecode (in ms) when this cluster ends. */
433	uint64_t tcEndMs;
434	/** Number of (simple) blocks in this cluster. */
435	uint64_t cBlocks;
436	};
437
438	/**
439	* Structure for keeping a WebM segment entry.
440	*
441	* Current we're only using one segment.
442	*/
443	struct WebMSegment
444	{
445	WebMSegment(void)
446	: tcStart(UINT64_MAX)
447	, tcEnd(UINT64_MAX)
448	, offStart(0)
449	, offInfo(0)
450	, offSeekInfo(0)
451	, offTracks(0)
452	, offCues(0)
453	{
454	uTimecodeScaleFactor = VBOX_WEBM_TIMECODESCALE_FACTOR_MS;
455
456	LogFunc(("Default timecode scale is: %RU64ns\n", uTimecodeScaleFactor));
457	}
458
459	/** The timecode scale factor of this segment. */
460	uint64_t uTimecodeScaleFactor;
461
462	/** Timecode when starting this segment. */
463	uint64_t tcStart;
464	/** Timecode when this segment ended. */
465	uint64_t tcEnd;
466
467	/** Absolute offset (in bytes) of CurSeg. */
468	uint64_t offStart;
469	/** Absolute offset (in bytes) of general info. */
470	uint64_t offInfo;
471	/** Absolute offset (in bytes) of seeking info. */
472	uint64_t offSeekInfo;
473	/** Absolute offset (in bytes) of tracks. */
474	uint64_t offTracks;
475	/** Absolute offset (in bytes) of cues table. */
476	uint64_t offCues;
477	/** List of cue points. Needed for seeking table. */
478	std::list<WebMCuePoint> lstCues;
479
480	/** Map of tracks.
481	* The key marks the track number (not the UUID!). */
482	std::map <uint8_t, WebMTrack *> mapTracks;
483
484	/** Current cluster which is being handled.
485	*
486	* Note that we don't need (and shouldn't need, as this can be a lot of data!) a
487	* list of all clusters. */
488	WebMCluster CurCluster;
489
490	} CurSeg;
491
492	#ifdef VBOX_WITH_LIBOPUS
493	# pragma pack(push)
494	# pragma pack(1)
495	/** Opus codec private data.
496	* Taken from: https://wiki.xiph.org/MatroskaOpus */
497	struct OpusPrivData
498	{
499	OpusPrivData(uint32_t a_u32SampleRate, uint8_t a_u8Channels)
500	: u8Channels(a_u8Channels)
501	, u32SampleRate(a_u32SampleRate) { }
502
503	/** "OpusHead". */
504	uint8_t au8Head[8] = { 0x4f, 0x70, 0x75, 0x73, 0x48, 0x65, 0x61, 0x64 };
505	/** Must set to 1. */
506	uint8_t u8Version = 1;
507	uint8_t u8Channels = 0;
508	uint16_t u16PreSkip = 0;
509	/** Sample rate before encoding to Opus.
510	* Note: This rate has nothing to do with the playback rate later! */
511	uint32_t u32SampleRate = 0;
512	uint16_t u16Gain = 0;
513	/** Must stay 0 -- otherwise a mapping table must be appended
514	* right after this header. */
515	uint8_t u8MappingFamily = 0;
516	};
517	# pragma pack(pop)
518	#endif /* VBOX_WITH_LIBOPUS */
519
520	/** Audio codec to use. */
521	WebMWriter::AudioCodec m_enmAudioCodec;
522	/** Video codec to use. */
523	WebMWriter::VideoCodec m_enmVideoCodec;
524
525	/** Whether we're currently in the tracks section. */
526	bool m_fInTracksSection;
527
528	/** Size of timecodes in bytes. */
529	size_t m_cbTimecode;
530	/** Maximum value a timecode can have. */
531	uint32_t m_uTimecodeMax;
532
533	Ebml m_Ebml;
534
535	public:
536
537	typedef std::map <uint8_t, WebMTrack *> WebMTracks;
538
539	public:
540
541	WebMWriter_Impl() :
542	m_fInTracksSection(false)
543	{
544	/* Size (in bytes) of time code to write. We use 2 bytes (16 bit) by default. */
545	m_cbTimecode = 2;
546	m_uTimecodeMax = UINT16_MAX;
547	}
548
549	virtual ~WebMWriter_Impl()
550	{
551	close();
552	}
553
554	/**
555	* Adds an audio track.
556	*
557	* @returns IPRT status code.
558	* @param uHz Input sampling rate.
559	* Must be supported by the selected audio codec.
560	* @param cChannels Number of input audio channels.
561	* @param cBits Number of input bits per channel.
562	* @param puTrack Track number on successful creation. Optional.
563	*/
564	int AddAudioTrack(uint16_t uHz, uint8_t cChannels, uint8_t cBits, uint8_t *puTrack)
565	{
566	#ifdef VBOX_WITH_LIBOPUS
567	int rc;
568
569	/*
570	* Check if the requested codec rate is supported.
571	*
572	* Only the following values are supported by an Opus standard build
573	* -- every other rate only is supported by a custom build.
574	*/
575	switch (uHz)
576	{
577	case 48000:
578	case 24000:
579	case 16000:
580	case 12000:
581	case 8000:
582	rc = VINF_SUCCESS;
583	break;
584
585	default:
586	rc = VERR_NOT_SUPPORTED;
587	break;
588	}
589
590	if (RT_FAILURE(rc))
591	return rc;
592
593	const uint8_t uTrack = (uint8_t)CurSeg.mapTracks.size();
594
595	m_Ebml.subStart(MkvElem_TrackEntry);
596
597	m_Ebml.serializeUnsignedInteger(MkvElem_TrackNumber, (uint8_t)uTrack);
598	m_Ebml.serializeString (MkvElem_Language, "und" /* "Undefined"; see ISO-639-2. */);
599	m_Ebml.serializeUnsignedInteger(MkvElem_FlagLacing, (uint8_t)0);
600
601	WebMTrack *pTrack = new WebMTrack(WebMTrackType_Audio, uTrack, RTFileTell(m_Ebml.getFile()));
602
603	pTrack->Audio.uHz = uHz;
604	pTrack->Audio.msPerBlock = 20; /** Opus uses 20ms by default. Make this configurable? */
605	pTrack->Audio.framesPerBlock = uHz / (1000 /* s in ms */ / pTrack->Audio.msPerBlock);
606
607	OpusPrivData opusPrivData(uHz, cChannels);
608
609	LogFunc(("Opus @ %RU16Hz (%RU16ms + %RU16 frames per block)\n",
610	pTrack->Audio.uHz, pTrack->Audio.msPerBlock, pTrack->Audio.framesPerBlock));
611
612	m_Ebml.serializeUnsignedInteger(MkvElem_TrackUID, pTrack->uUUID, 4)
613	.serializeUnsignedInteger(MkvElem_TrackType, 2 /* Audio */)
614	.serializeString(MkvElem_CodecID, "A_OPUS")
615	.serializeData(MkvElem_CodecPrivate, &opusPrivData, sizeof(opusPrivData))
616	.serializeUnsignedInteger(MkvElem_CodecDelay, 0)
617	.serializeUnsignedInteger(MkvElem_SeekPreRoll, 80 * 1000000) /* 80ms in ns. */
618	.subStart(MkvElem_Audio)
619	.serializeFloat(MkvElem_SamplingFrequency, (float)uHz)
620	.serializeUnsignedInteger(MkvElem_Channels, cChannels)
621	.serializeUnsignedInteger(MkvElem_BitDepth, cBits)
622	.subEnd(MkvElem_Audio)
623	.subEnd(MkvElem_TrackEntry);
624
625	CurSeg.mapTracks[uTrack] = pTrack;
626
627	if (puTrack)
628	*puTrack = uTrack;
629
630	return VINF_SUCCESS;
631	#else
632	RT_NOREF(uHz, cChannels, cBits, puTrack);
633	return VERR_NOT_SUPPORTED;
634	#endif
635	}
636
637	int AddVideoTrack(uint16_t uWidth, uint16_t uHeight, double dbFPS, uint8_t *puTrack)
638	{
639	#ifdef VBOX_WITH_LIBVPX
640	RT_NOREF(dbFPS);
641
642	const uint8_t uTrack = (uint8_t)CurSeg.mapTracks.size();
643
644	m_Ebml.subStart(MkvElem_TrackEntry);
645
646	m_Ebml.serializeUnsignedInteger(MkvElem_TrackNumber, (uint8_t)uTrack);
647	m_Ebml.serializeString (MkvElem_Language, "und" /* "Undefined"; see ISO-639-2. */);
648	m_Ebml.serializeUnsignedInteger(MkvElem_FlagLacing, (uint8_t)0);
649
650	WebMTrack *pTrack = new WebMTrack(WebMTrackType_Video, uTrack, RTFileTell(m_Ebml.getFile()));
651
652	/** @todo Resolve codec type. */
653	m_Ebml.serializeUnsignedInteger(MkvElem_TrackUID, pTrack->uUUID /* UID */, 4)
654	.serializeUnsignedInteger(MkvElem_TrackType, 1 /* Video */)
655	.serializeString(MkvElem_CodecID, "V_VP8")
656	.subStart(MkvElem_Video)
657	.serializeUnsignedInteger(MkvElem_PixelWidth, uWidth)
658	.serializeUnsignedInteger(MkvElem_PixelHeight, uHeight)
659	.subEnd(MkvElem_Video);
660
661	m_Ebml.subEnd(MkvElem_TrackEntry);
662
663	CurSeg.mapTracks[uTrack] = pTrack;
664
665	if (puTrack)
666	*puTrack = uTrack;
667
668	return VINF_SUCCESS;
669	#else
670	RT_NOREF(uWidth, uHeight, dbFPS, puTrack);
671	return VERR_NOT_SUPPORTED;
672	#endif
673	}
674
675	int writeHeader(void)
676	{
677	LogFunc(("Header @ %RU64\n", RTFileTell(m_Ebml.getFile())));
678
679	m_Ebml.subStart(MkvElem_EBML)
680	.serializeUnsignedInteger(MkvElem_EBMLVersion, 1)
681	.serializeUnsignedInteger(MkvElem_EBMLReadVersion, 1)
682	.serializeUnsignedInteger(MkvElem_EBMLMaxIDLength, 4)
683	.serializeUnsignedInteger(MkvElem_EBMLMaxSizeLength, 8)
684	.serializeString(MkvElem_DocType, "webm")
685	.serializeUnsignedInteger(MkvElem_DocTypeVersion, 2)
686	.serializeUnsignedInteger(MkvElem_DocTypeReadVersion, 2)
687	.subEnd(MkvElem_EBML);
688
689	m_Ebml.subStart(MkvElem_Segment);
690
691	/* Save offset of current segment. */
692	CurSeg.offStart = RTFileTell(m_Ebml.getFile());
693
694	writeSegSeekInfo();
695
696	/* Save offset of upcoming tracks segment. */
697	CurSeg.offTracks = RTFileTell(m_Ebml.getFile());
698
699	/* The tracks segment starts right after this header. */
700	m_Ebml.subStart(MkvElem_Tracks);
701	m_fInTracksSection = true;
702
703	return VINF_SUCCESS;
704	}
705
706	int writeSimpleBlockInternal(WebMTrack *a_pTrack, uint64_t a_uTimecode,
707	const void *a_pvData, size_t a_cbData, uint8_t a_fFlags)
708	{
709	Log3Func(("SimpleBlock @ %RU64 (T%RU8, TS=%RU64, %zu bytes)\n",
710	RTFileTell(m_Ebml.getFile()), a_pTrack->uTrack, a_uTimecode, a_cbData));
711
712	/** @todo Mask out non-valid timecode bits, e.g. the upper 48 bits for a (default) 16-bit timecode. */
713	Assert(a_uTimecode <= m_uTimecodeMax);
714
715	/* Write a "Simple Block". */
716	m_Ebml.writeClassId(MkvElem_SimpleBlock);
717	/* Block size. */
718	m_Ebml.writeUnsignedInteger(0x10000000u \| ( 1 /* Track number size. */
719	+ m_cbTimecode /* Timecode size .*/
720	+ 1 /* Flags size. */
721	+ a_cbData /* Actual frame data size. */), 4);
722	/* Track number. */
723	m_Ebml.writeSize(a_pTrack->uTrack);
724	/* Timecode (relative to cluster opening timecode). */
725	m_Ebml.writeUnsignedInteger(a_uTimecode, m_cbTimecode);
726	/* Flags. */
727	m_Ebml.writeUnsignedInteger(a_fFlags, 1);
728	/* Frame data. */
729	m_Ebml.write(a_pvData, a_cbData);
730
731	a_pTrack->cTotalBlocks++;
732
733	return VINF_SUCCESS;
734	}
735
736	#ifdef VBOX_WITH_LIBVPX
737	int writeBlockVP8(WebMTrack a_pTrack, const vpx_codec_enc_cfg_t a_pCfg, const vpx_codec_cx_pkt_t *a_pPkt)
738	{
739	RT_NOREF(a_pTrack);
740
741	WebMCluster &Cluster = CurSeg.CurCluster;
742
743	/* Calculate the PTS of this frame (in ms). */
744	uint64_t tcPTSMs = a_pPkt->data.frame.pts * 1000
745	* (uint64_t) a_pCfg->g_timebase.num / a_pCfg->g_timebase.den;
746
747	if ( tcPTSMs
748	&& tcPTSMs <= CurSeg.CurCluster.tcEndMs)
749	{
750	tcPTSMs = CurSeg.CurCluster.tcEndMs + 1;
751	}
752
753	/* Whether to start a new cluster or not. */
754	bool fClusterStart = false;
755
756	/* No blocks written yet? Start a new cluster. */
757	if (a_pTrack->cTotalBlocks == 0)
758	fClusterStart = true;
759
760	/* Did we reach the maximum a cluster can hold? Use a new cluster then. */
761	if (tcPTSMs > VBOX_WEBM_CLUSTER_MAX_LEN_MS)
762	{
763	tcPTSMs = 0;
764
765	fClusterStart = true;
766	}
767
768	const bool fKeyframe = RT_BOOL(a_pPkt->data.frame.flags & VPX_FRAME_IS_KEY);
769
770	if ( fClusterStart
771	\|\| fKeyframe)
772	{
773	if (Cluster.fOpen) /* Close current cluster first. */
774	{
775	m_Ebml.subEnd(MkvElem_Cluster);
776	Cluster.fOpen = false;
777	}
778
779	Cluster.fOpen = true;
780	Cluster.uID = a_pTrack->cTotalClusters;
781	Cluster.tcStartMs = tcPTSMs;
782	Cluster.offCluster = RTFileTell(m_Ebml.getFile());
783	Cluster.cBlocks = 0;
784
785	LogFunc(("[T%RU8C%RU64] Start @ %RU64ms / %RU64 bytes\n",
786	a_pTrack->uTrack, Cluster.uID, Cluster.tcStartMs, Cluster.offCluster));
787
788	m_Ebml.subStart(MkvElem_Cluster)
789	.serializeUnsignedInteger(MkvElem_Timecode, Cluster.tcStartMs);
790
791	/* Save a cue point if this is a keyframe. */
792	if (fKeyframe)
793	{
794	WebMCuePoint cue(a_pTrack, Cluster.tcStartMs, Cluster.offCluster);
795	CurSeg.lstCues.push_back(cue);
796	}
797
798	a_pTrack->cTotalClusters++;
799	}
800
801	Cluster.tcEndMs = tcPTSMs;
802	Cluster.cBlocks++;
803
804	/* Calculate the block's timecode, which is relative to the cluster's starting timecode. */
805	uint16_t tcBlockMs = static_cast<uint16_t>(tcPTSMs - Cluster.tcStartMs);
806
807	Log2Func(("tcPTSMs=%RU64, tcBlockMs=%RU64\n", tcPTSMs, tcBlockMs));
808
809	Log2Func(("[C%RU64] cBlocks=%RU64, tcStartMs=%RU64, tcEndMs=%RU64 (%zums)\n",
810	Cluster.uID, Cluster.cBlocks, Cluster.tcStartMs, Cluster.tcEndMs, Cluster.tcEndMs - Cluster.tcStartMs));
811
812	uint8_t fFlags = 0;
813	if (fKeyframe)
814	fFlags \|= VBOX_WEBM_BLOCK_FLAG_KEY_FRAME;
815	if (a_pPkt->data.frame.flags & VPX_FRAME_IS_INVISIBLE)
816	fFlags \|= VBOX_WEBM_BLOCK_FLAG_INVISIBLE;
817
818	return writeSimpleBlockInternal(a_pTrack, tcBlockMs, a_pPkt->data.frame.buf, a_pPkt->data.frame.sz, fFlags);
819	}
820	#endif /* VBOX_WITH_LIBVPX */
821
822	#ifdef VBOX_WITH_LIBOPUS
823	/* Audio blocks that have same absolute timecode as video blocks SHOULD be written before the video blocks. */
824	int writeBlockOpus(WebMTrack a_pTrack, const void pvData, size_t cbData, uint64_t uTimeStampMs)
825	{
826	AssertPtrReturn(a_pTrack, VERR_INVALID_POINTER);
827	AssertPtrReturn(pvData, VERR_INVALID_POINTER);
828	AssertReturn(cbData, VERR_INVALID_PARAMETER);
829
830	RT_NOREF(uTimeStampMs);
831
832	WebMCluster &Cluster = CurSeg.CurCluster;
833
834	/* Calculate the PTS of the current block:
835	*
836	* The "raw PTS" is the exact time of an object represented in nanoseconds):
837	* Raw Timecode = (Block timecode + Cluster timecode) * TimecodeScaleFactor
838	*/
839	uint64_t tcPTSMs = Cluster.tcStartMs + (Cluster.cBlocks * 20 /ms /);
840
841	/* Whether to start a new cluster or not. */
842	bool fClusterStart = false;
843
844	if (a_pTrack->cTotalBlocks == 0)
845	fClusterStart = true;
846
847	/* Did we reach the maximum a cluster can hold? Use a new cluster then. */
848	if (tcPTSMs > VBOX_WEBM_CLUSTER_MAX_LEN_MS)
849	fClusterStart = true;
850
851	if (fClusterStart)
852	{
853	if (Cluster.fOpen) /* Close current clusters first. */
854	{
855	m_Ebml.subEnd(MkvElem_Cluster);
856	Cluster.fOpen = false;
857	}
858
859	Cluster.fOpen = true;
860	Cluster.uID = a_pTrack->cTotalClusters;
861	Cluster.tcStartMs = Cluster.tcEndMs;
862	Cluster.offCluster = RTFileTell(m_Ebml.getFile());
863	Cluster.cBlocks = 0;
864
865	LogFunc(("[T%RU8C%RU64] Start @ %RU64ms / %RU64 bytes\n",
866	a_pTrack->uTrack, Cluster.uID, Cluster.tcStartMs, Cluster.offCluster));
867
868	/* As all audio frame as key frames, insert a new cue point when a new cluster starts. */
869	WebMCuePoint cue(a_pTrack, Cluster.tcStartMs, Cluster.offCluster);
870	CurSeg.lstCues.push_back(cue);
871
872	m_Ebml.subStart(MkvElem_Cluster)
873	.serializeUnsignedInteger(MkvElem_Timecode, Cluster.tcStartMs);
874
875	a_pTrack->cTotalClusters++;
876	}
877
878	Cluster.tcEndMs = tcPTSMs;
879	Cluster.cBlocks++;
880
881	/* Calculate the block's timecode, which is relative to the cluster's starting timecode. */
882	uint16_t tcBlockMs = static_cast<uint16_t>(tcPTSMs - Cluster.tcStartMs);
883
884	Log2Func(("tcPTSMs=%RU64, tcBlockMs=%RU64\n", tcPTSMs, tcBlockMs));
885
886	Log2Func(("[C%RU64] cBlocks=%RU64, tcStartMs=%RU64, tcEndMs=%RU64 (%zums)\n",
887	Cluster.uID, Cluster.cBlocks, Cluster.tcStartMs, Cluster.tcEndMs, Cluster.tcEndMs - Cluster.tcStartMs));
888
889	return writeSimpleBlockInternal(a_pTrack, tcBlockMs,
890	pvData, cbData, VBOX_WEBM_BLOCK_FLAG_KEY_FRAME);
891	}
892	#endif /* VBOX_WITH_LIBOPUS */
893
894	int WriteBlock(uint8_t uTrack, const void *pvData, size_t cbData)
895	{
896	RT_NOREF(pvData, cbData); /* Only needed for assertions for now. */
897
898	WebMTracks::iterator itTrack = CurSeg.mapTracks.find(uTrack);
899	if (itTrack == CurSeg.mapTracks.end())
900	return VERR_NOT_FOUND;
901
902	WebMTrack *pTrack = itTrack->second;
903	AssertPtr(pTrack);
904
905	int rc;
906
907	if (m_fInTracksSection)
908	{
909	m_Ebml.subEnd(MkvElem_Tracks);
910	m_fInTracksSection = false;
911	}
912
913	switch (pTrack->enmType)
914	{
915
916	case WebMTrackType_Audio:
917	{
918	#ifdef VBOX_WITH_LIBOPUS
919	if (m_enmAudioCodec == WebMWriter::AudioCodec_Opus)
920	{
921	Assert(cbData == sizeof(WebMWriter::BlockData_Opus));
922	WebMWriter::BlockData_Opus pData = (WebMWriter::BlockData_Opus )pvData;
923	rc = writeBlockOpus(pTrack, pData->pvData, pData->cbData, pData->uTimestampMs);
924	}
925	else
926	#endif /* VBOX_WITH_LIBOPUS */
927	rc = VERR_NOT_SUPPORTED;
928	break;
929	}
930
931	case WebMTrackType_Video:
932	{
933	#ifdef VBOX_WITH_LIBVPX
934	if (m_enmVideoCodec == WebMWriter::VideoCodec_VP8)
935	{
936	Assert(cbData == sizeof(WebMWriter::BlockData_VP8));
937	WebMWriter::BlockData_VP8 pData = (WebMWriter::BlockData_VP8 )pvData;
938	rc = writeBlockVP8(pTrack, pData->pCfg, pData->pPkt);
939	}
940	else
941	#endif /* VBOX_WITH_LIBVPX */
942	rc = VERR_NOT_SUPPORTED;
943	break;
944	}
945
946	default:
947	rc = VERR_NOT_SUPPORTED;
948	break;
949	}
950
951	return rc;
952	}
953
954	int writeFooter(void)
955	{
956	if (m_fInTracksSection)
957	{
958	m_Ebml.subEnd(MkvElem_Tracks);
959	m_fInTracksSection = false;
960	}
961
962	if (CurSeg.CurCluster.fOpen)
963	{
964	m_Ebml.subEnd(MkvElem_Cluster);
965	CurSeg.CurCluster.fOpen = false;
966	}
967
968	/*
969	* Write Cues element.
970	*/
971	LogFunc(("Cues @ %RU64\n", RTFileTell(m_Ebml.getFile())));
972
973	CurSeg.offCues = RTFileTell(m_Ebml.getFile());
974
975	m_Ebml.subStart(MkvElem_Cues);
976
977	std::list<WebMCuePoint>::iterator itCuePoint = CurSeg.lstCues.begin();
978	while (itCuePoint != CurSeg.lstCues.end())
979	{
980	/* Sanity. */
981	AssertPtr(itCuePoint->pTrack);
982
983	const uint64_t uClusterPos = itCuePoint->offClusterStart - CurSeg.offStart;
984
985	LogFunc(("CuePoint @ %RU64: Track #%RU8 (Time %RU64, Pos %RU64)\n",
986	RTFileTell(m_Ebml.getFile()), itCuePoint->pTrack->uTrack, itCuePoint->tcClusterStart, uClusterPos));
987
988	m_Ebml.subStart(MkvElem_CuePoint)
989	.serializeUnsignedInteger(MkvElem_CueTime, itCuePoint->tcClusterStart)
990	.subStart(MkvElem_CueTrackPositions)
991	.serializeUnsignedInteger(MkvElem_CueTrack, itCuePoint->pTrack->uTrack)
992	.serializeUnsignedInteger(MkvElem_CueClusterPosition, uClusterPos, 8)
993	.subEnd(MkvElem_CueTrackPositions)
994	.subEnd(MkvElem_CuePoint);
995
996	itCuePoint++;
997	}
998
999	m_Ebml.subEnd(MkvElem_Cues);
1000	m_Ebml.subEnd(MkvElem_Segment);
1001
1002	/*
1003	* Re-Update SeekHead / Info elements.
1004	*/
1005
1006	writeSegSeekInfo();
1007
1008	return RTFileSeek(m_Ebml.getFile(), 0, RTFILE_SEEK_END, NULL);
1009	}
1010
1011	int close(void)
1012	{
1013	WebMTracks::iterator itTrack = CurSeg.mapTracks.begin();
1014	for (; itTrack != CurSeg.mapTracks.end(); ++itTrack)
1015	{
1016	delete itTrack->second;
1017	CurSeg.mapTracks.erase(itTrack);
1018	}
1019
1020	Assert(CurSeg.mapTracks.size() == 0);
1021
1022	m_Ebml.close();
1023
1024	return VINF_SUCCESS;
1025	}
1026
1027	friend class Ebml;
1028	friend class WebMWriter;
1029
1030	private:
1031
1032	/**
1033	* Writes the segment's seek information and cue points.
1034	*
1035	* @returns IPRT status code.
1036	*/
1037	void writeSegSeekInfo(void)
1038	{
1039	if (CurSeg.offSeekInfo)
1040	RTFileSeek(m_Ebml.getFile(), CurSeg.offSeekInfo, RTFILE_SEEK_BEGIN, NULL);
1041	else
1042	CurSeg.offSeekInfo = RTFileTell(m_Ebml.getFile());
1043
1044	LogFunc(("SeekHead @ %RU64\n", CurSeg.offSeekInfo));
1045
1046	m_Ebml.subStart(MkvElem_SeekHead);
1047
1048	m_Ebml.subStart(MkvElem_Seek)
1049	.serializeUnsignedInteger(MkvElem_SeekID, MkvElem_Tracks)
1050	.serializeUnsignedInteger(MkvElem_SeekPosition, CurSeg.offTracks - CurSeg.offStart, 8)
1051	.subEnd(MkvElem_Seek);
1052
1053	Assert(CurSeg.offCues - CurSeg.offStart > 0); /* Sanity. */
1054
1055	m_Ebml.subStart(MkvElem_Seek)
1056	.serializeUnsignedInteger(MkvElem_SeekID, MkvElem_Cues)
1057	.serializeUnsignedInteger(MkvElem_SeekPosition, CurSeg.offCues - CurSeg.offStart, 8)
1058	.subEnd(MkvElem_Seek);
1059
1060	m_Ebml.subStart(MkvElem_Seek)
1061	.serializeUnsignedInteger(MkvElem_SeekID, MkvElem_Info)
1062	.serializeUnsignedInteger(MkvElem_SeekPosition, CurSeg.offInfo - CurSeg.offStart, 8)
1063	.subEnd(MkvElem_Seek);
1064
1065	m_Ebml.subEnd(MkvElem_SeekHead);
1066
1067	//int64_t iFrameTime = (int64_t)1000 * 1 / 25; //m_Framerate.den / m_Framerate.num; /** @todo Fix this! */
1068	CurSeg.offInfo = RTFileTell(m_Ebml.getFile());
1069
1070	LogFunc(("Info @ %RU64\n", CurSeg.offInfo));
1071
1072	char szMux[64];
1073	RTStrPrintf(szMux, sizeof(szMux),
1074	#ifdef VBOX_WITH_LIBVPX
1075	"vpxenc%s", vpx_codec_version_str());
1076	#else
1077	"unknown");
1078	#endif
1079	char szApp[64];
1080	RTStrPrintf(szApp, sizeof(szApp), VBOX_PRODUCT " %sr%u", VBOX_VERSION_STRING, RTBldCfgRevision());
1081
1082	LogFunc(("Duration=%RU64\n", CurSeg.tcEnd - CurSeg.tcStart));
1083
1084	m_Ebml.subStart(MkvElem_Info)
1085	.serializeUnsignedInteger(MkvElem_TimecodeScale, CurSeg.uTimecodeScaleFactor)
1086	.serializeFloat(MkvElem_Segment_Duration, CurSeg.tcEnd - CurSeg.tcStart)
1087	.serializeString(MkvElem_MuxingApp, szMux)
1088	.serializeString(MkvElem_WritingApp, szApp)
1089	.subEnd(MkvElem_Info);
1090	}
1091	};
1092
1093	WebMWriter::WebMWriter(void) : m_pImpl(new WebMWriter_Impl()) {}
1094
1095	WebMWriter::~WebMWriter(void)
1096	{
1097	if (m_pImpl)
1098	{
1099	Close();
1100	delete m_pImpl;
1101	}
1102	}
1103
1104	int WebMWriter::Create(const char *a_pszFilename, uint64_t a_fOpen,
1105	WebMWriter::AudioCodec a_enmAudioCodec, WebMWriter::VideoCodec a_enmVideoCodec)
1106	{
1107	try
1108	{
1109	m_pImpl->m_enmAudioCodec = a_enmAudioCodec;
1110	m_pImpl->m_enmVideoCodec = a_enmVideoCodec;
1111
1112	LogFunc(("Creating '%s'\n", a_pszFilename));
1113
1114	int rc = m_pImpl->m_Ebml.create(a_pszFilename, a_fOpen);
1115	if (RT_SUCCESS(rc))
1116	rc = m_pImpl->writeHeader();
1117	}
1118	catch(int rc)
1119	{
1120	return rc;
1121	}
1122	return VINF_SUCCESS;
1123	}
1124
1125	int WebMWriter::Close(void)
1126	{
1127	if (!m_pImpl->m_Ebml.isOpen())
1128	return VINF_SUCCESS;
1129
1130	int rc = m_pImpl->writeFooter();
1131	if (RT_SUCCESS(rc))
1132	m_pImpl->close();
1133
1134	return rc;
1135	}
1136
1137	int WebMWriter::AddAudioTrack(uint16_t uHz, uint8_t cChannels, uint8_t cBitDepth, uint8_t *puTrack)
1138	{
1139	return m_pImpl->AddAudioTrack(uHz, cChannels, cBitDepth, puTrack);
1140	}
1141
1142	int WebMWriter::AddVideoTrack(uint16_t uWidth, uint16_t uHeight, double dbFPS, uint8_t *puTrack)
1143	{
1144	return m_pImpl->AddVideoTrack(uWidth, uHeight, dbFPS, puTrack);
1145	}
1146
1147	int WebMWriter::WriteBlock(uint8_t uTrack, const void *pvData, size_t cbData)
1148	{
1149	int rc;
1150
1151	try
1152	{
1153	rc = m_pImpl->WriteBlock(uTrack, pvData, cbData);
1154	}
1155	catch(int rc2)
1156	{
1157	rc = rc2;
1158	}
1159	return rc;
1160	}
1161
1162	const Utf8Str& WebMWriter::GetFileName(void)
1163	{
1164	return m_pImpl->m_Ebml.getFileName();
1165	}
1166
1167	uint64_t WebMWriter::GetFileSize(void)
1168	{
1169	return m_pImpl->m_Ebml.getFileSize();
1170	}
1171
1172	uint64_t WebMWriter::GetAvailableSpace(void)
1173	{
1174	return m_pImpl->m_Ebml.getAvailableSpace();
1175	}
1176

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/VBox/Main/src-client/EbmlWriter.cpp@ 68451

Download in other formats: