SoC 2010 Ogg Demuxer - Revision history

DoesItReallyMatter: Tag with {{SoCProject}} (sorry if this pings anyone)

2019-02-05T07:08:40Z

Tag with {{SoCProject}} (sorry if this pings anyone)

← Older revision		Revision as of 07:08, 5 February 2019
Line 1:		Line 1:
			{{SoCProject\|year=2010\|student=[[User:salsaman\|salsaman]]\|mentor=Christophe Mutricy}}

	'''Week 1:'''		'''Week 1:'''

J-b at 13:45, 30 September 2013

2013-09-30T13:45:22Z

← Older revision		Revision as of 13:45, 30 September 2013
Line 365:		Line 365:

	c) VLC_CAN_EXACTSEEK flag for stream would be nice (see above).		c) VLC_CAN_EXACTSEEK flag for stream would be nice (see above).

			{{GSoC}}

Salsaman at 02:37, 10 August 2010

2010-08-10T02:37:42Z

← Older revision		Revision as of 02:37, 10 August 2010
Line 344:		Line 344:


			Summary

	'''~~Week 7+:~~'''		'''9th August'''
			Most of the rest of the time has been spent transforming my updates into a sequence of patches (9 in all), and testing the results.

	~~TBD. (Maybe look at x and y offsets, maybe look at multiple clips within one ogg container, possibly add ogg skeleton support).~~		There are still a couple of outstanding issues to be resolved:

	~~Unofficial codecs~~ in ogg ~~(ogm, e~~.g. ~~PCM audio, divx, xvid) ?~~		a) The function to get the length (time) of a clip has been implemented in ogg.c. If there is a theora or dirac video stream present then the precise video duration will be returned. If neither of these streams are present then VLC_EGENERIC is returned.

	~~Frei0r effects ?~~		It seems that enabling this function and returning VLC_EGENERIC causes a regression whereby the estimated duration is no longer set.

	~~Jack audio output ?~~		There are two parts to fixing this:
			1) vlc itself needs to check for VLC_EGENERIC returned from DEMUX_GET_LENGTH and do whatever calculation it was doing previously when this function was not implemented.

	~~dv demuxer~~ / ~~decoder ?~~		2) due to an oversight on my part I had not considered audio only ogg/vorbis files. For these files there is a method to get an accurate length. I have agreed with the vlc team to implement this after the end of the GSOC project (probably as an update in september).

	~~Port other demuxers/decoders to LiVES ?~~

			b) seeking in dirac still requires communcation of timestamps between demuxer and codec. It may not be optimal to implement this since it creates a strong coupling between the demuxer and decoder plugin. However if such a shared data scheme is implemented then ogg/dirac seeking can be properly performed. I have given an example implementation of this using block->next, but there may be better solutions to be devised.


	~~'''Note:'''~~		c) VLC_CAN_EXACTSEEK flag for stream would be nice (see above).

	~~I will~~ be ~~mostly unavailable after 29th July~~ (~~attending Piksel Summer Camp and then family commitments~~)~~, but I should have all of the coding for the main project done well before that.~~

	~~'''Update'''~~

	~~I have amended my travel plans and will now be available until 10th August~~.

Salsaman at 14:47, 14 July 2010

2010-07-14T14:47:41Z

← Older revision		Revision as of 14:47, 14 July 2010
Line 313:		Line 313:
	OK, I am almost done with porting the dirac seeking code into vlc. However there is one big problem - when the decoder decodes a frame it needs to pass the frame number back to the demuxer. However there seems to be no mechanism for doing this. My first idea was to set a dummy block as the block->p_next and have the decoder set the i_pts field for this, however this does not work as vlc tries to parse block->p_next as an actual block and frees it.		OK, I am almost done with porting the dirac seeking code into vlc. However there is one big problem - when the decoder decodes a frame it needs to pass the frame number back to the demuxer. However there seems to be no mechanism for doing this. My first idea was to set a dummy block as the block->p_next and have the decoder set the i_pts field for this, however this does not work as vlc tries to parse block->p_next as an actual block and frees it.

	~~'''I have sent a couple of emails to my mentor asking for suggestions, but so far there has been no response.~~
	~~'''~~

			''For now I cannot continue any further with this part of the project (dirac seeking).''
	''For now I cannot continue any further with this part of the project.
	''


Line 329:		Line 325:
	This behaviour can be avoided by uncommenting #define DISABLE_DIRAC_SEEK in ogg.h, which will revert to the old style of handling dirac.		This behaviour can be avoided by uncommenting #define DISABLE_DIRAC_SEEK in ogg.h, which will revert to the old style of handling dirac.


			'''Weeks 7 & 8'''


			'''14 Jul'''
			I have been spending the last few days reviewing my work so far (thanks to some feedback from mentors). At their suggestion I have been working on splitting all my contributions into 4 patches:

			- non-seek related logic and code refactoring
			- seek logic changes
			- theora seek capability
			- dirac seek capability

			These patches can then be applied in sequence. I have completed a first pass of this, now I need to tweak the patches a little.


Line 359:		Line 364:

	I will be mostly unavailable after 29th July (attending Piksel Summer Camp and then family commitments), but I should have all of the coding for the main project done well before that.		I will be mostly unavailable after 29th July (attending Piksel Summer Camp and then family commitments), but I should have all of the coding for the main project done well before that.

			'''Update'''

			I have amended my travel plans and will now be available until 10th August.

Salsaman at 06:16, 2 July 2010

2010-07-02T06:16:28Z

← Older revision		Revision as of 06:16, 2 July 2010
Line 320:		Line 320:
	''		''


			'''2 Jul'''


			I have submitted a patch with everything I have done so far. The patch suffers from the p_next problem noted yesterday and will crash in block_fifoEmpty. This seems to be an error in vlc, as I have set the free function,
			block->pf_free to a function which does nothing.

			This behaviour can be avoided by uncommenting #define DISABLE_DIRAC_SEEK in ogg.h, which will revert to the old style of handling dirac.

Salsaman at 18:37, 1 July 2010

2010-07-01T18:37:08Z

← Older revision		Revision as of 18:37, 1 July 2010
Line 291:		Line 291:

	'''Update''': I can now get the last granulepos (and hence number of frames) for even the smallest of files. Unfortunately this caused some breakage elsewhere, so I will have to do a bit of debugging.		'''Update''': I can now get the last granulepos (and hence number of frames) for even the smallest of files. Unfortunately this caused some breakage elsewhere, so I will have to do a bit of debugging.






Line 301:		Line 305:

	Today I began backporting the new search code frome LiVES back into vlc.		Today I began backporting the new search code frome LiVES back into vlc.




			'''1 Jul'''

			OK, I am almost done with porting the dirac seeking code into vlc. However there is one big problem - when the decoder decodes a frame it needs to pass the frame number back to the demuxer. However there seems to be no mechanism for doing this. My first idea was to set a dummy block as the block->p_next and have the decoder set the i_pts field for this, however this does not work as vlc tries to parse block->p_next as an actual block and frees it.

			'''I have sent a couple of emails to my mentor asking for suggestions, but so far there has been no response.
			'''


			''For now I cannot continue any further with this part of the project.
			''

Salsaman at 04:10, 29 June 2010

2010-06-29T04:10:56Z

← Older revision		Revision as of 04:10, 29 June 2010
Line 298:		Line 298:
	'''TODO''': finish debugging regression, port code into vlc.		'''TODO''': finish debugging regression, port code into vlc.

			'''29 Jun'''

			Today I began backporting the new search code frome LiVES back into vlc.

Salsaman at 23:30, 25 June 2010

2010-06-25T23:30:46Z

← Older revision		Revision as of 23:30, 25 June 2010
Line 290:		Line 290:
	Still working on some minor code corrections. Frame count for very small ogg (and maybe dirac) files is still broken; the last granulepos is not being found.		Still working on some minor code corrections. Frame count for very small ogg (and maybe dirac) files is still broken; the last granulepos is not being found.

	Update: I can now get the last granulepos (and hence number of frames) for even the smallest of files. Unfortunately this caused some breakage elsewhere, so I will have to do a bit of debugging.		'''Update''': I can now get the last granulepos (and hence number of frames) for even the smallest of files. Unfortunately this caused some breakage elsewhere, so I will have to do a bit of debugging.


Line 296:		Line 296:
	'''Week 6'''		'''Week 6'''

	TODO: finish debugging regression, port code into vlc.		'''TODO''': finish debugging regression, port code into vlc.

Salsaman at 23:29, 25 June 2010

2010-06-25T23:29:40Z

← Older revision		Revision as of 23:29, 25 June 2010
Line 290:		Line 290:
	Still working on some minor code corrections. Frame count for very small ogg (and maybe dirac) files is still broken; the last granulepos is not being found.		Still working on some minor code corrections. Frame count for very small ogg (and maybe dirac) files is still broken; the last granulepos is not being found.

			Update: I can now get the last granulepos (and hence number of frames) for even the smallest of files. Unfortunately this caused some breakage elsewhere, so I will have to do a bit of debugging.


Line 295:		Line 296:
	'''Week 6'''		'''Week 6'''

	~~Finish up dirac and theora seeking~~, port code into vlc.		TODO: finish debugging regression, port code into vlc.

	~~Still TODO: frame count for very small files.~~

Salsaman at 20:47, 25 June 2010

2010-06-25T20:47:53Z

← Older revision		Revision as of 20:47, 25 June 2010
Line 286:		Line 286:


			'''25 Jun'''

			Still working on some minor code corrections. Frame count for very small ogg (and maybe dirac) files is still broken; the last granulepos is not being found.




	'''Week 6+:'''		'''Week 6'''

			Finish up dirac and theora seeking, port code into vlc.

			Still TODO: frame count for very small files.






			'''Week 7+:'''

	TBD. (Maybe look at x and y offsets, maybe look at multiple clips within one ogg container, possibly add ogg skeleton support).		TBD. (Maybe look at x and y offsets, maybe look at multiple clips within one ogg container, possibly add ogg skeleton support).