[FFmpeg-devel-irc] IRC log for 2010-01-28

irc at mansr.com irc at mansr.com
Fri Jan 29 01:00:19 CET 2010


[18:37:27] * /join #ffmpeg-devel ...
[18:37:29] *** TOPIC: Welcome to the FFmpeg development channel. That is, development of FFmpeg, not using FFmpeg, nor libav*. | Users should redirect their questions to #ffmpeg | FFmpeg 0.5 has been released!
[18:37:29] *** TOPICINFO: superdump, 1236951942
[18:38:05] <mru> there, that should do it
[18:38:23] <mru> logs are sent to ffmpeg-devel-irc ML
[18:42:25] <kshishkov> mru: what do you think of Gdium?
[18:43:52] * kshishkov begins to suspect it's not worse than other netbooks
[18:51:25] <mru> about as useless as other netbooks
[18:51:44] <kshishkov> even a bit more
[18:52:01] <kshishkov> but compared to Acer crap it's a bit sturdier
[18:52:08] <mru> I haven't tried using it as intended
[18:52:15] <kshishkov> and a bit better display even 16-bit onyl
[18:52:36] <mru> my 4-year old laptop is better in every respect
[18:52:52] <mru> but then it was 4x the price too
[18:53:13] <kshishkov> the best laptop I worked on was some restored Fujitsu-Siemens owned by Benjamin
[18:54:08] <Compn> some of those old laptops, when restored really purr
[18:54:22] <Compn> compared to the overheating pos on todays market
[18:54:44] * kshishkov owns refurbished IBM TP390
[18:55:01] <kshishkov> but in that case it had been restored by Benjamin himself
[18:55:02] <Compn> loud fans, heat problems, dust buildup, hardware (especially wifi) failures
[18:55:17] <jai> the panasonic cf-r8 was quite good
[19:09:26] <bilboed> Dark_Shikari, what's the link of that david attenborough h264 you linked to earlier here ?
[19:09:51] <Dark_Shikari>  http://mirror05.x264.nl/Dark/x264clips/PlanetEarthBirds.mkv
[19:10:00] <Dark_Shikari> that's a clip from a rip (not the original blu-ray) of Planet Earth
[19:10:06] <Dark_Shikari> I highly recommend you watch the entire series.  It's awesome.
[19:10:14] <bilboed> the private life of birds ?
[19:10:19] <bilboed> already seen it :)
[19:10:32] <Dark_Shikari> the birds segment is the hardest to compress portion of it
[19:10:44] <Dark_Shikari> aka 3 million birds on screen at once
[19:10:51] <bilboed> most of Attenborough's stuff is friggin awesome tbh
[19:11:11] <Dark_Shikari> planet earth in HD was amazing
[19:11:43] * kshishkov is stuck with HD = everything > 352x288
[19:12:00] <twnqx> :S
[19:22:30] <bilboed> Dark_Shikari, ouch... yah, home computer can't cope with it (AMD Athlon(tm) 64 X2 Dual Core Processor 4600+)
[19:25:05] <kshishkov> wait for hw Theora decoders then ;)
[19:25:28] <bilboed> that's h264
[19:25:39] <kshishkov> I know
[19:26:05] <kshishkov> but you can actually buy and use hw H.264 decoder
[19:27:12] <kshishkov> try that with 1920x1080 Theora though
[19:27:29] <thresh> are there any dumbnuts who encode HD with Theora?
[19:27:35] <mru> xiph?
[19:27:50] <thresh> well, except :)
[19:28:07] <mru> hi DonDiego
[19:28:40] <kshishkov> maybe we'll get XiphTube eventually with full Firefox HTML5 support
[19:29:00] <DonDiego> http://lwn.net/Articles/370985/
[19:29:12] <jai> and worthless content to top it off
[19:29:25] <DonDiego> me flaming around h.264 and firefox...
[19:29:32] <DonDiego> and theora of course..
[19:30:15] <iive> I guess you are done with your exams.
[19:31:27] <Compn> btw i made the mistake of trying to install silverlight on win2k today
[19:31:54] <mru> were there any survivors?
[19:32:05] <Compn> well, i gotta reboot that box to make sure nothing died
[19:32:07] * Compn crosses fingers
[19:32:12] <iive> your computer committes suicide?
[19:33:13] <DonDiego> i'm especially fond of the part where i accuse fluendo of being tasteless and insulting
[19:34:06] <DonDiego> for claiming that they are the only ones who provide a "legal properly licensed" h.264 decoder
[19:34:33] <bilboed> I guess they don't know about coreavc
[19:35:10] <BBB> why bother about a company providing closed-source technology?
[19:35:19] <BBB> at least google provides an opensource, well-licensed h264 decoder
[19:35:34] <DonDiego> they often get portrayed as free sw champions
[19:35:41] <DonDiego> and seem to think so of themselves
[19:35:45] <BBB> the power of PR
[19:35:46] <bilboed> DonDiego, bullshit
[19:35:53] <bilboed> I'm sure BBB will chime in here :)
[19:35:59] <BBB> huh?
[19:36:06] <BBB> I have no idea what you're talking about
[19:36:07] <kshishkov> same for On2
[19:36:08] <BBB> (really)
[19:36:36] <DonDiego> who is thomasvs then?
[19:36:48] <DonDiego> http://lwn.net/Articles/371647/
[19:37:03] <DonDiego> (notice that i didn't start flinging the word tasteless around)
[19:37:35] <DonDiego> bilboed: so fluendo knows they're an enemy of freedom like any other?
[19:37:48] <bilboed> DonDiego, they're a business
[19:37:49] <bilboed> full stop
[19:37:55] <bilboed> they used to sponsor gstreamer development
[19:37:59] <bilboed> that stopped 3 years ago
[19:38:03] <DonDiego> oh
[19:38:06] <DonDiego> interesting
[19:38:09] <bilboed> https://www.ohloh.net/p/gstreamer/contributors
[19:38:12] <DonDiego> tell me more about it please
[19:38:22] <kshishkov> is sponsoring gstreamer _that_ evil?
[19:38:22] <bilboed> look at thomasvs' commit graph
[19:38:26] <bilboed> I'm not inventing anything
[19:38:58] <BBB> holy shit, I'm in that list
[19:39:00] <BBB> funny
[19:39:02] <bilboed> BBB, yah :)
[19:39:43] * bilboed carries on reading comments
[19:40:07] <bilboed> kshishkov, I'd say no... but then we are one of the main companies sponsoring now, so slightly biased
[19:40:19] <DonDiego> we?
[19:40:35] <bilboed> Collabora Multimedia
[19:40:49] <mru> what a silly name
[19:40:50] <bilboed> sorry, should have put that in the sentence :(
[19:44:16] <bilboed> mru, eh, luckily most people don't give a damn about names, they care about results
[19:44:42] <mru> I just think it sounds silly
[19:44:47] <mru> sorry if that offends you
[19:44:59] <bilboed> it doesn't
[19:45:04] <CIA-27> ffmpeg: michael * r21514 /trunk/libavcodec/h264_loopfilter.c:
[19:45:04] <CIA-27> ffmpeg: factorize first filter call out, this makes the code somewhat
[19:45:04] <CIA-27> ffmpeg: smaller without any speed loss.
[19:47:11] <bilboed> DonDiego, didn't you ask for the h264 license agreement (the document that is, to read) ?
[19:47:22] <DonDiego> i found it
[19:47:25] <bilboed> oh
[19:47:30] <bilboed> you got a link ?
[19:47:35] <DonDiego> and i'm getting it fedexed to my house as well
[19:47:47] <DonDiego> i just sent an email to the mpeg la
[19:47:59] <Dark_Shikari> bah, I'm feeling stupid today
[19:48:00] <bilboed> they're sending it free of charge ?
[19:48:00] <Dark_Shikari> $ ffmpeg -y -i 00000.m2ts -vcodec copy -acodec copy -map 0.0 -map 0.1 output.mk
[19:48:03] <Dark_Shikari> v
[19:48:09] <Dark_Shikari> this is trying to copy all the streams, not just 0.0 and 0.1
[19:48:14] <Dark_Shikari> how am I failing at using map?
[19:48:28] <DonDiego> bilboed: yes, of course
[19:48:35] <bilboed> DonDiego, interesting
[19:48:54] <drv> Dark_Shikari: isn't it 0:1, not 0.1?
[19:49:07] <Dark_Shikari> still doesn't work
[19:49:10] <DonDiego> bilboed: umm, do you expect them to charge for the contract papers?
[19:49:13] <Dark_Shikari> "number of stream maps must match number of output streams"
[19:49:14] <DonDiego> bilboed: http://apps.shareholder.com/sec/viewerContent.aspx?companyid=DIVX&docid=6470121
[19:49:24] <bilboed> DonDiego, good point
[19:49:35] <DonDiego> bilboed: scroll way down and look for the exhibits
[19:49:40] <Dark_Shikari> oh, I had to use -sn
[19:49:50] <J_Darnley> Dark_Shikari: subtitles?
[19:49:54] <J_Darnley> ...
[19:50:34] <bilboed> DonDiego, AVC Patent Portfolio License Agreement ?
[19:50:41] <DonDiego> yes
[19:50:42] <Dark_Shikari> bilboed: I have that on my desk :)
[19:50:57] <Dark_Shikari> (if you need it for any reference
[19:51:38] <bilboed> sure, let me just grab a ticket to the west coast pronto :)
[19:51:51] <Dark_Shikari> I mean if you want me to look at something ;)
[19:52:01] <DonDiego> wasn't there somebody working for mpeg la in this channel at some point?
[19:52:04] <Compn> DonDiego : btw, where did you get the idea to argue theora vs h264 ?
[19:52:28] <KotH> BBB, mru: pong
[19:52:34] <DonDiego> i just cannot let a good flaming opportunity pass me by ;-p
[19:52:56] <mru> we have an aliasing bug that actually breaks stuff :-(
[19:52:59] <mru> somewhere in h264
[19:53:17] <Dark_Shikari> Then go apply those god damned x264 aliasing macros
[19:53:23] <Dark_Shikari> type-puns are not tolerable anymore
[19:54:34] <KotH> BBB: tell me what dns entries you want, who manages the registration, etc pp and i do the rest
[19:54:35] <mru> some of the avc tests are failing on blackfin
[19:54:48] <mru> if I compile h264*.o with -fno-strict-aliasing they pass
[19:54:55] <KotH> BBB: if you are next week at fosdem and you are not in a hurry, we can do it there
[19:55:24] <Dark_Shikari> mru: then apply the damned macros.
[19:55:35] <mru> I'm not going to take that fight
[19:55:38] <Dark_Shikari> what fight?
[19:55:40] <Dark_Shikari> it has no impact on speed
[19:55:45] <Dark_Shikari> if anything some tests were faster after I used them in x264
[19:55:46] <mru> the fight to edit h264*.c
[19:55:51] <Dark_Shikari> I don't mean h264.c
[19:55:53] <Dark_Shikari> I mean all files in ffmpeg
[19:56:02] <Dark_Shikari> find and replace every instance of *(uint32_t*)
[19:56:04] <Dark_Shikari> and similar
[19:56:07] * KotH goes back into hiding....
[19:56:13] <mru> that's what I'm not going to fight about
[19:56:20] <BBB> KotH: sorry, too far away
[19:56:28] <Dark_Shikari> well, until that's fixed, not a line of code in ffmpeg can be trusted
[19:56:31] <kshishkov> IIRC there were some in rv34.c
[19:56:42] <BBB> koth: can we do it now? I have the domains already
[19:56:43] <Dark_Shikari> same with [0][x+y*stride] addressing of arrays
[19:56:54] <Dark_Shikari> this is not helping ffmpeg's reputation
[19:56:57] <kshishkov> Dark_Shikari: I usually trust empty and commented out lines
[19:57:00] <mru> that's been slammed by michael already
[19:57:03] <mru> he refuses to fix it
[19:57:20] <Dark_Shikari> Then find a nice way to fix it, like we did in x264
[19:57:25] <Dark_Shikari> we changed all [8][8] DCT arrays to [64]
[19:57:30] <Dark_Shikari> that had zero speed cost and fixed the problem
[19:57:43] <Dark_Shikari> And just commit it
[19:58:28] <bilboed> DonDiego, ah, one thing that would be interesting is to figure out whether the license is per unit *sold* (stressing the sold point)
[19:58:42] <bilboed> DonDiego, license cost that is
[19:58:52] <DonDiego> it says so in the summary
[19:58:59] <DonDiego> i haven't read the complete license yet
[19:59:04] <Dark_Shikari> I have the license, I can check
[19:59:07] <bilboed> DonDiego, yah, I meant in the fine prints
[19:59:09] <bilboed> Dark_Shikari, yay
[19:59:24] <Compn> heh
[19:59:29] <mru> http://thread.gmane.org/gmane.comp.video.ffmpeg.devel/61948
[19:59:33] <bilboed> because firefox is free afaik :)
[19:59:37] * Compn looks at epic opera vs google flame
[19:59:57] <Dark_Shikari> bilboed: also, it depends on country
[20:00:00] <bilboed> Compn, what ? there's *another* flame ? :)
[20:00:05] <Dark_Shikari> AFAIK, in europe you can distribute any software without paying
[20:00:07] <bilboed> Dark_Shikari, aaaah, so basically the US are fucked
[20:00:08] <Compn> bilboed : just the same html5 one
[20:00:09] <Dark_Shikari> HOWEVER, if you combine the software with hardware
[20:00:14] <Dark_Shikari> you have to pay
[20:00:24] <Dark_Shikari> because patents on "computer programs" are legally prohibited
[20:00:25] <bilboed> Compn, I imagined :) although I'm guessing gstreamer was thrown in the mix
[20:00:34] <DonDiego> what's the status of opera and html5?
[20:00:37] <bilboed> Dark_Shikari, ... in europe, right ?
[20:00:51] <Compn> have to read opera blog for that i think
[20:00:56] <Compn> or ask in #opera
[20:01:01] <Compn> forgot what network that was on
[20:01:02] <Dark_Shikari> bilboed: yes
[20:01:03] <bilboed> DonDiego, gst-based, it'll pick up any gstreamer plugin you have on your system (on linux), therefore works with gst-ffmpeg... and any codec :)
[20:01:07] <Dark_Shikari> just going by european union law
[20:01:09] <Dark_Shikari> ok, I got it
[20:01:27] <Dark_Shikari> "hereof upon the Sale after Dec. 31, 2004, of each AVC Encoder, each AVC Decoder, or each AVC Codec"
[20:01:29] <DonDiego> Dark_Shikari: in our dreams, they are prohibited, in practice the european patent office hands out sw patents aplenty
[20:01:33] <Dark_Shikari> let's see where it defines "sale"
[20:01:46] <Dark_Shikari> DonDiego: yes, the reason being that they apply to the mix of software + hardware
[20:02:01] <Dark_Shikari> Oops.  Sorry bilboed.  No dice.
[20:02:18] <bilboed> Dark_Shikari, let me guess : sale == distribution even for free ? :)
[20:02:28] <Dark_Shikari> "Sale: shall mean any sale, rental, lease, license, copying, transfer, reproduction, transmission, or other form of distribution of an AVC Product or the Transmission by any means of AVC Video either directly or through a chain of distribution."
[20:02:37] <bilboed> fuckers
[20:02:43] <bilboed> *sigh*
[20:02:58] <bilboed> DonDiego, see why I wanted the fine prints ? :)
[20:03:28] <DonDiego> yes, you always need to look at the fine print, *sigh*
[20:04:09] <DonDiego> Dark_Shikari: no hw necessary to get a patent here, pure sw is getting patented all the time
[20:04:23] <mru> as someone said, the large print giveth, and the small print taketh away
[20:04:24] <Dark_Shikari> DonDiego: computer software is counted as "hardware" if it's put on a computer ;)
[20:04:35] <DonDiego> the fuckers have no shame and violate both the letter and the spirit of the european patent agreement every day
[20:04:37] <bilboed> DonDiego, we tried (many many many many many many times) to convince mozilla to use gstreamer for many reasons, but also for one that might interest you : they wouldn't have to ship h264... but it could be one click away for the user (install gst-ffmpeg).
[20:04:43] <Dark_Shikari> DonDiego: then they'll lose in court
[20:04:52] <mru> not if they buy the judge
[20:04:58] <bilboed> DonDiego, which would bring it to the same situation as linux distros right now. They don't *ship* ffmpeg... but it's a click away to install
[20:05:03] <Dark_Shikari> bilboed: exactly
[20:05:05] <Dark_Shikari> they don't care about patents
[20:05:10] <Dark_Shikari> they won't support anything except theora and vorbis
[20:05:10] <Dark_Shikari> not dirac
[20:05:11] <Dark_Shikari> not speex
[20:05:13] <Dark_Shikari> not flac
[20:05:16] <DonDiego> it's what's going to come..
[20:05:26] <Dark_Shikari> they engineered their entire video layer to be unable to work with anything else
[20:05:31] <bilboed> eh, even the N900 has ffmpeg one click away now :)
[20:05:48] <DonDiego> a firefox extension/plugin/whatever or fork that retrofits h.264 support
[20:05:56] <Dark_Shikari> yup, it will
[20:06:01] <Dark_Shikari> once youtube rolls out html5 officially it will come
[20:06:11] <Dark_Shikari> and then drama will occur
[20:06:12] <bilboed> DonDiego, you want the best part ? .... they use gstreamer for fennec (their embedded version of firefox)
[20:06:19] <bilboed> they already have the code !!!oneone111
[20:06:28] <DonDiego> i still wonder what internet explorer will do..
[20:06:32] <bilboed> Dark_Shikari, uncertain
[20:06:38] <DonDiego> bilboed: then please join the flames in that thread
[20:06:39] <BBB> I think you're overhyping gstreamer, bilboed
[20:06:40] <bilboed> Dark_Shikari, a lot of people will stick to firefox I think
[20:06:45] <BBB> similar to that chrome bug report
[20:06:50] <DonDiego> roc from mozilla is claiming the opposite
[20:06:51] <BBB> many projects don't *want* that complexity
[20:07:02] <Dark_Shikari> DonDiego: IE will support it using Windows 7 Media Foundation
[20:07:10] <DonDiego> yes?
[20:07:13] <DonDiego> what is that?
[20:07:21] <Dark_Shikari> It's the new directshow-like system
[20:07:21] <bilboed> BBB, I agree to a certain extent... but ... you end up with a system where one click away you can install ffmpeg
[20:07:30] <bilboed> Dark_Shikari, oh... blow ... me
[20:07:35] <Dark_Shikari> It supports h264 out of the box
[20:07:37] <BBB> bilboed, I didn't say ffmpeg was the solution
[20:07:40] <DonDiego> mru: please don't shrink away from the aliasing flames^wdiscussion
[20:08:18] <bilboed> BBB, hey, you didn't come to OVC last june in NYC ? you would have loved the html5 roundtable :)
[20:08:38] <BBB> no time, have to study
[20:08:41] <BBB> will come some other time
[20:08:49] <bilboed> hopefully they do it again this year
[20:09:15] <bilboed> Dark_Shikari, yah, h264 on win7 is a smart move. was about time
[20:09:37] <bilboed> although I wonder how bad their decoder is :(
[20:10:00] <Dark_Shikari> it's pretty bad
[20:10:19] <Dark_Shikari> http://img64.imageshack.us/img64/4789/comparison2.png
[20:10:29] <mru> hmm, I'm still getting 4 failures with h264 built w/o strict aliasing
[20:10:31] <DonDiego> but ie 8 does not support html5 right now, right?
[20:10:36] <bilboed> no
[20:10:46] <Dark_Shikari> IE9 probably will.
[20:10:46] <DonDiego> when will they release the next version?
[20:10:46] <mru> was 11 with standard build
[20:11:11] <bilboed> Dark_Shikari, 2160p => the ms decoder can't even play it ?
[20:11:11] <elenril> what makes coreavc 2 so much faster?
[20:11:16] <Dark_Shikari> correct
[20:11:23] <bilboed> Dark_Shikari, ... lol ? :)
[20:11:24] <Dark_Shikari> coreavc 1.9.5 also had an arbitrary width limit of 2048
[20:11:32] <Dark_Shikari> and cores to 4
[20:11:37] <Dark_Shikari> which is why it was so slow in tests
[20:12:27] <bilboed> 2160p divx ?
[20:12:44] <Dark_Shikari> just an upscaled 1080p clip at low bitrate
[20:13:21] <DonDiego> bilboed: will you comment on lwn.net?
[20:13:35] <bilboed> DonDiego, I'm a read-only user of lwn :)
[20:13:53] <DonDiego> well, get yourself an account :)
[20:13:56] <elenril> Shipping known-patented code without a license is tasteless << lol
[20:14:10] <bilboed> DonDiego, no, I have an account
[20:14:26] <bilboed> DonDiego, commenting on comments just pisses me off
[20:14:32] <DonDiego> why?
[20:14:37] <Compn> elenril : thats one hell of a quote :\
[20:14:37] <CIA-27> ffmpeg: mru * r21515 /trunk/common.mak:
[20:14:37] <CIA-27> ffmpeg: Append ECFLAGS to CFLAGS
[20:14:37] <CIA-27> ffmpeg: This makes it easy to supply extra flags on the make command line
[20:14:37] <CIA-27> ffmpeg: for testing purposes.
[20:14:49] <Compn> DonDiego : no reason to debate with idiots :P
[20:14:50] <Dark_Shikari> spambots are back.
[20:14:51] <bilboed> DonDiego, I'll quote a famous person "I'm getting too old for this shit"
[20:14:56] <Compn> hehe
[20:15:07] <DonDiego> i can understand that sentiment, but
[20:15:15] <Compn> btw whats with gstreamer joining in adding to fud ?
[20:15:16] <bilboed> Compn, yah, you could translate it to that too :)
[20:15:17] <DonDiego> discussions on lwn.net are not /.
[20:15:24] <Dark_Shikari> Compn: because they are intent on being our enemy really
[20:15:30] <bilboed> Compn, who from gstreamer commented in that thread ?
[20:15:31] <Dark_Shikari> which is unfortunate
[20:15:41] <DonDiego> it's high-quality arguments and the real developers exchanging arguments
[20:15:41] <Compn> bilboed : someone linked elsewhere with gstreamer quotes
[20:15:45] <bilboed> Dark_Shikari, bullshit, and you know it very well
[20:15:51] <Dark_Shikari> gstreamer wouldn't really exist without ffmpeg
[20:15:53] <Compn> lets see if i can dig it up
[20:15:54] <Dark_Shikari> and yet gstreamer constantly flames ffmpeg
[20:16:00] <Dark_Shikari> now, of course, everyone flames ffmpeg
[20:16:02] <Dark_Shikari> because it's fucking fun
[20:16:05] <bilboed> Dark_Shikari, works perfectly fine on many devices without ffmpeg :)
[20:16:09] <Dark_Shikari> but you have to stop at some point ;)
[20:16:13] <Dark_Shikari> bilboed: sure, with proprietary codecs
[20:16:24] <Compn> bilboed : http://gstreamer.freedesktop.org/data/doc/gstreamer/head/faq/html/chapter-legal.html#legal-distribute-three
[20:16:28] <bilboed> Dark_Shikari, sure, we don't care whether it's proprietary of open soruce codecs
[20:16:39] <DonDiego> and that is the problem..
[20:16:41] <Dark_Shikari> bilboed: which makes the idea of gstreamer going full-freetard hilarious
[20:16:49] <Dark_Shikari> and yeah, that is a problem IMO
[20:16:55] <DonDiego> bilboed: anyway, will you mention that fennec uses gstreamer there?
[20:17:05] <DonDiego> otherwise please provide me with some suitable links
[20:17:26] <bilboed> DonDiego, let me dig that up
[20:17:41] <Dark_Shikari> bilboed: it's just that I don't like when people try to position themselves as opponents of ffmpeg despite not actually being an opponent
[20:18:03] <bilboed> ffs... I'm *not* an opponent of ffmpeg
[20:18:04] <Dark_Shikari> I know
[20:18:06] <Dark_Shikari> not talking about you
[20:18:18] <Dark_Shikari> you != all gstreamer people ;)
[20:18:24] <Compn> bilboed : oh nevermind, i think it was an idiot who quoted that page
[20:18:25] <bilboed> and I'm pretty sure a *lot* of gst devs and users are all for ffmpeg
[20:18:28] <Dark_Shikari> I know
[20:18:31] <Compn> bilboed : please ignore my comments about gst
[20:18:38] <Dark_Shikari> oh, good.
[20:18:40] <Compn> what Dark_Shikari has against gst i dunno
[20:18:43] <Compn> :)
[20:18:48] <bilboed> Compn, distribution is a bitch tbh
[20:18:52] <Dark_Shikari> against gstreamer?  I find the interface really annoying
[20:19:01] <Dark_Shikari> and last time I tried to do something really simple it wasn't capable of it
[20:19:04] <Dark_Shikari> (remux from mkv to TS)
[20:19:14] <bilboed> Dark_Shikari, we suck at high-level stuff (i.e. user-friendly)
[20:19:17] <bilboed> Dark_Shikari, and we know it
[20:19:24] <bilboed> Dark_Shikari, I'm partly trying to solve this
[20:19:25] <Dark_Shikari> bilboed: well then you're in good company!
[20:19:26] <Dark_Shikari> <rimshot>
[20:19:36] <Dark_Shikari> well, except ffmpeg isn't trying to solve it.
[20:19:40] <Dark_Shikari> <double rimshot?>
[20:19:43] <bilboed> rofl
[20:19:58] <Compn> video encoding is not user-friendly
[20:20:00] <mru> Dark_Shikari: when did gcc get the may_alias attribute?
[20:20:01] <Compn> :P
[20:20:05] <Dark_Shikari> mru: ages ago iirc
[20:20:14] <Dark_Shikari> pengvado would know
[20:20:29] <Dark_Shikari> I think we added it under gcc 3.3 or later
[20:20:43] <BBB> koth is gone again :(
[20:20:55] <Dark_Shikari> yes
[20:20:57] <Dark_Shikari> 3.3 and later.
[20:20:59] <Dark_Shikari> confirmed by gcc docs.
[20:21:42] <mru> or I could just test for it
[20:21:50] <mru> we shouldn't need to use it in public headers anyway
[20:22:35] <Dark_Shikari> of course
[20:25:35] <bilboed> DonDiego, fyi, this is the gst-backend for mozilla bug report where they mentionn it's used for fennec : https://bugzilla.mozilla.org/show_bug.cgi?id=422540
[20:26:18] <Compn> Dark_Shikari : btw, some info on the freendoe spam >>> http://unethicalblogger.com/posts/2010/01/using_browser_piss_irc_users_or_spamming_redditdowntime
[20:26:38] <Compn> in short: its all reddits fault.
[20:26:50] <Dark_Shikari> lol
[20:26:50] <bilboed> DonDiego, warning though, that bug report is massive :(
[21:13:12] <ohsix> DonDiego: re: that lwn thread, doesn't the bit about "Sales" posted from the actual license conflict with what you kept saying
[21:13:41] <DonDiego> probably
[21:13:55] <DonDiego> i'd have to reread the patent license in detail
[21:14:47] <ohsix> well their definition of sale includes any sort of distribution among other things
[21:18:30] <DonDiego> apparently
[21:18:42] <DonDiego> but then there's their definition of intermediate product or so
[21:19:02] <DonDiego> i remember that from some other pdf or ppt they had lying around
[22:07:14] <Compn> i dont know how mike runs the apple swf blog
[22:14:19] <Yuvi> speaking of theora vs. h264, http://codecism.blogspot.com/
[22:16:26] <mru> btw, the bbc pretty much canned dirac development because h264 was so cheap they didn't need their own
[22:16:50] <Yuvi> yeah, libdirac hasn't seen a single commit in nearly a year
[22:17:39] <ShadowJK> Oh youtube does h264 baseline? Is it properly marked as such?
[22:21:39] <DonDiego> whose blog is that?
[22:22:03] <mru> a good guy, apparently
[22:35:28] <astrange> youtube is baseline+8x8dct = high
[22:35:53] <astrange> i don't think it uses any other features, but i only looked with -debug 1
[22:53:51] <ohsix> Yuvi: heh that dude used Canaan as one of his samples :> thats what i'm hitting bugs in gst-ffmpeg with
[22:58:00] <mru> Dark_Shikari: do you know what code in ffh264 deals with frext stuff?
[22:58:20] <Dark_Shikari> which frext stuff ?
[22:59:12] <mru> http://pastebin.ca/1769979
[22:59:14] <mru> those samples
[22:59:27] <Dark_Shikari> that question doesn't make sense
[22:59:30] <Dark_Shikari> "deals with frext stuff"
[22:59:42] <Dark_Shikari> almost all the code "deals with" those samples
[22:59:58] <mru> yes, but those use something that the others don't
[23:00:02] <mru> and it's breaking
[23:00:14] <Dark_Shikari> let me see
[23:00:22] <mru> I remember frext support being added
[23:00:41] <mru> but I don't remember what it is
[23:01:21] <Dark_Shikari> it's called "high profile"
[23:01:23] <Dark_Shikari> >_>
[23:01:32] <mru> part of high profile
[23:01:37] <Dark_Shikari> no, it IS high profile
[23:01:44] <Dark_Shikari> "fidelity range extensions" was the original name of high profile
[23:01:48] <mru> not every HP stream has to use that stuff
[23:01:55] <mru> oh
[23:02:39] <Dark_Shikari> Streameye says that first one is corrupt
[23:02:53] <Dark_Shikari> though I don't trust it
[23:02:59] <mru> doesn't matter
[23:03:03] <Dark_Shikari> it has a CQM
[23:03:12] <mru> I still get different output depending on strict aliasing on or off
[23:04:31] <Dark_Shikari> is either of those correct outputs?
[23:04:40] <mru> yes
[23:04:50] <mru> according to the md5sums at leasst
[23:04:55] <Dark_Shikari> which one is correct?
[23:05:02] <mru> -fno-strict-aliasing
[23:05:03] <mru> of course
[23:05:18] <mru> but I can't find any more dodgy pointer stuff
[23:05:27] <mru> all the warnings are gone
[23:05:28] <Dark_Shikari> Does the first I-frame match?
[23:06:43] <Dark_Shikari> md5 wise
[23:06:59] <mru> no
[23:07:48] <Dark_Shikari> they all seem to have CQMs
[23:10:02] <Dark_Shikari> the frext panasonic ones don't
[23:10:16] <Dark_Shikari> lol ,frext 3 breaks CMA1820 analyzer
[23:10:29] <Dark_Shikari> oh god at frext 4, mixed mbaff/paff
[23:28:40] <mru> Dark_Shikari: know any tricks for finding aliasing violations?
[23:28:47] <mru> the ones gcc doesn't warn about
[23:28:52] <Dark_Shikari> remove them all
[23:28:59] <Dark_Shikari> and it will warn about them all
[23:29:06] <mru> no it doesn't
[23:29:06] <Dark_Shikari> gcc 4.4 warns about things that are not even aliasing violations
[23:29:10] <mru> I'm using 4.4
[23:29:11] <Dark_Shikari> it is absurdly overzealous
[23:29:14] <Dark_Shikari> it will whine about every last type pun
[23:29:26] <mru> no warnings, yet -fno-strict-aliasing makes a difference
[23:29:49] <Dark_Shikari> no warnings? wtf?
[23:29:53] <Dark_Shikari> h264 is covered with type puns
[23:29:58] <mru> I removed them all
[23:30:02] <Dark_Shikari> o.0
[23:30:03] <mru> the ones it warned about
[23:30:14] <Dark_Shikari> Did you remove all cases of [0][X+y*stride]?
[23:30:18] <Dark_Shikari> no, you have to remove them all
[23:30:28] <Dark_Shikari> again, the only proper way to do it is to remove ALL VIOLATIONS
[23:30:30] <mru> does it warn about those?
[23:30:31] <Dark_Shikari> otherwise the code cannot be trusted
[23:30:32] <Dark_Shikari> no it doesn't
[23:30:39] <mru> there we go then...
[23:30:53] <BBB> it couldn't, the code might be intended that way
[23:31:06] <mru> it bloody well shouldn't be
[23:31:26] <Dark_Shikari> intentional array overflows are dangerous
[23:31:29] <BBB> bloody is about as british as a curse can be
[23:31:49] <mru> british curses are the best
[23:33:53] <Dark_Shikari> mru: it would have been easier to add aliasing-avoidance macros
[23:33:55] <Dark_Shikari> than to remove all the type puns
[23:33:58] <Dark_Shikari> you could have already done it
[23:34:04] <mru> I did
[23:34:14] <Dark_Shikari> ah
[23:40:10] <pengvado> gcc warns about [0][X+y*stride] if it can prove that X+y*stride will exceed the minor dimension. but if they're variables then it doesn't know.
[23:43:32] <Dark_Shikari> yup
[23:48:47] <mru> h264.h:1220
[23:48:50] <mru> what's that doing?
[23:49:46] <Dark_Shikari> converting reference frame index to picture index
[23:50:32] <mru> what's with the +20
[23:50:38] <Dark_Shikari> No idea
[23:50:56] <Dark_Shikari> I don't know shit about mbaff deblocking rules
[23:50:58] <Dark_Shikari> and I don't plan to learn
[23:50:58] <mru> gcc complains
[23:51:35] <Dark_Shikari> about which part
[23:52:01] <mru> libavcodec/h264.h:1221: warning: initialization from incompatible pointer type
[23:52:43] <mru> that would be 1220 in svn sources


More information about the FFmpeg-devel-irc mailing list