[Ffmpeg-devel-irc] ffmpeg-devel.log.20160423

burek burek021 at gmail.com
Sun Apr 24 02:05:02 CEST 2016


[00:03:51 CEST] <BBB> super-creepy
[00:03:53 CEST] <BBB> ohwell
[00:03:59 CEST] <BBB> I dont see how my thing fixes anything though
[00:04:10 CEST] <BBB> it allows conversion
[00:04:13 CEST] <BBB> this is more like input
[00:04:32 CEST] <JEEB> + cross-library dependencies
[00:04:57 CEST] <JEEB> well basically I would like it if getting r,g,b or y,cb,cr values for one another would be centralized somewhere
[00:05:43 CEST] <JEEB> together with general conversion functionality for planes
[01:18:56 CEST] <BBB> JEEB: but most of it is much more purpose-specific than that
[01:19:14 CEST] <BBB> JEEB: e.g. in vf_colorspace, we convert to intermediates that are 15bit with limited overflow (but not absent)
[01:19:22 CEST] <BBB> JEEB: swscale has different rules
[01:19:30 CEST] <BBB> JEEB: and typcailly youd want clipped 8bit to 8bit
[01:19:33 CEST] <BBB> so its all different
[03:20:02 CEST] <cone-768> ffmpeg 03Michael Niedermayer 07master:a39e3e5e7b12: avutil/aes: Remove duplicate include
[11:13:26 CEST] <cone-659> ffmpeg 03Carl Eugen Hoyos 07master:0499012a5592: lavf/isom: Support Perian tag NELL for nellymoser.
[13:46:34 CEST] <wm4> wat...
[13:48:38 CEST] <fritsch> hehe
[14:17:54 CEST] <rcombs> ubitux: http://puu.sh/osIs4/c7dd3182b9.aac <-- does this look like AAC to you
[14:18:39 CEST] <rcombs> erm actually http://puu.sh/osIv2/29b9046e18.aac
[14:23:23 CEST] <ubitux> why do you ask me? i don't know much about aac
[14:23:34 CEST] <ubitux> close to nothing tbh
[14:24:50 CEST] <rcombs> uh I was aiming for atomnuker and failed spectacularly
[14:25:02 CEST] <rcombs> this is a good indicator that it's time for me to get some sleep
[14:27:24 CEST] <atomnuker> rcombs: looks fairly high entropy, might be quantized coefficients, doesn't look like huffman codes from scalefactors
[14:28:12 CEST] <rcombs> lavc doesn't like it, but I saw what looked very much like ADTS frame headers
[14:29:19 CEST] <rcombs> (playing around with Amazon Video)
[14:30:50 CEST] <atomnuker> you'd think they'd use encryption, wouldn't they?
[14:31:37 CEST] <rcombs> this is post-encryption, I _think_
[14:31:43 CEST] <rcombs> erm, decryption
[14:32:18 CEST] <rcombs> (set breakpoint on decryption function; let function run; examine contents of buffers)
[14:33:48 CEST] <atomnuker> they have to decoder the data somewhere too, and chances are they'll send full packets there
[14:34:12 CEST] <rcombs> Chrome provides two paths; one where the decryption plug-in also decodes, and one where it does not
[14:34:17 CEST] <rcombs> this takes the latter path
[14:35:40 CEST] <rcombs> doesn't help that I'm looking at a stripped Chrome binary and poking around in memory mostly-blind
[14:37:15 CEST] <wm4> can't you build a non-stripped one?
[14:37:23 CEST] <rcombs> I might try
[14:37:34 CEST] <wm4> just make sure you have gigabytes of memory
[14:37:34 CEST] <rcombs> unfortunately chromium is a behemoth of misery
[14:37:39 CEST] <JEEB> can chromium even use the drm libraries for chrome?
[14:37:45 CEST] <rcombs> I'd imagine so
[14:37:51 CEST] <JEEB> ok
[14:38:02 CEST] <rcombs> like, I don't see a _technical_ reason it couldn't
[14:38:08 CEST] <JEEB> and yes, building chromium is a very... *special* effort
[14:38:13 CEST] <rcombs> they're just PPAPI plugins
[14:38:42 CEST] <rcombs> they run 'em out-of-process and pass "decrypt" messages around
[14:38:43 CEST] <JEEB> at my previous job it was rather simple to trigger a xwalk build, which contained chromium :)
[14:39:01 CEST] <JEEB> and then you'd go for a long coffee break
[14:39:13 CEST] <JEEB> (when you updated your own middleware component)
[14:40:11 CEST] <rcombs> I was really looking for video and got sidetracked when I saw something that looked vaguely like AAC
[14:40:29 CEST] <rcombs> and manually parsed the ADTS header and went "!!!!!" because it looked sane
[14:40:34 CEST] <JEEB> :D
[14:40:38 CEST] <rcombs> (stereo, 48kHz&)
[14:40:39 CEST] <JEEB> I know that feeling
[14:40:59 CEST] <JEEB> in the end the last time I poked a DRM scheme I ended up just hooking ReadFile that the player was using
[14:41:09 CEST] <JEEB> (yes, it was the shitty type of DRM where you have to install a rootkit player)
[14:41:25 CEST] <rcombs> I was hoping to trap the VideoToolbox decode call
[14:41:41 CEST] <rcombs> but apparently they're not using that
[14:41:57 CEST] <rcombs> so& I guess they're using software decoding at expense of battery life and performance
[14:42:02 CEST] <rcombs> for the sake of CONTENT SECURITY
[14:42:05 CEST] <JEEB> yup
[14:42:26 CEST] <JEEB> and video is shown through the "secure route" available on macs and windows, amirite?
[14:42:27 CEST] <rcombs> also I guess they must pay for the codec licenses then
[14:42:32 CEST] Action: rcombs shrugs
[14:42:47 CEST] <JEEB> yes, google already pretty much pays the maximum amount
[14:42:50 CEST] <JEEB> so they don't really care
[14:42:51 CEST] <JEEB> :)
[14:43:16 CEST] <rcombs> I just took a screenshot and it came out normal
[14:43:20 CEST] <JEEB> löl
[14:43:30 CEST] <rcombs> so& I dunno whether or not "secure route" means that shouldn't work
[14:43:38 CEST] <JEEB> it should be a black box with the secure route
[14:43:43 CEST] <JEEB> so yeah, they're not even doing that :)
[14:44:28 CEST] <JEEB> I think netflix's implementation tried to use that one? never tried though
[14:45:24 CEST] <rcombs> also the screen recording feature in QuickTime works
[14:45:28 CEST] <wm4> and at least win d3d11 api is full of stuff for the secure route
[14:45:53 CEST] <wm4> keep in mind that DRM is for restricting the _consumer_, not pirates
[14:46:08 CEST] <JEEB> yup
[14:46:22 CEST] <rcombs> <insert xkcd link>
[14:46:26 CEST] <JEEB> and for various legal reasons
[14:46:27 CEST] <wm4> but if a simple screenrecorder works, then lol
[14:46:45 CEST] <JEEB> wm4: well they're not using the secure route in the amazon case it seems
[14:47:30 CEST] <rcombs> this is really more of a challenge than anything practical
[14:48:09 CEST] <rcombs> I guess since there must be an H.264 decoder in this binary somewhere, I could just& look for it, and trap on that
[14:48:22 CEST] <rcombs> can't be _that_ hard to find, looking at a mountain of disassembly
[14:49:52 CEST] <wm4> at some point they got to output video
[14:50:11 CEST] <rcombs> make it only work on the built-in display
[14:50:14 CEST] <wm4> knowing nothing about anything, this sounds like the easiest point of attack
[14:50:17 CEST] <rcombs> (am I evil enough)
[14:50:46 CEST] <rcombs> wm4: yeah but then you have to re-encode :P
[15:32:06 CEST] <JEEB> can someone confirm what the default value is for codec context's height f.ex.?
[15:32:11 CEST] <JEEB> -1 or 0
[15:38:20 CEST] <wm4> just check <= 0
[15:39:52 CEST] <JEEB> k
[16:42:08 CEST] <nevcairiel> rcombs: chromium can most definitely use the widevine drm module from chrome, and even play amazon video with it
[16:42:36 CEST] <nevcairiel> i set that up a couple weeks back, only took b uilding my own chromium to enable it
[17:36:30 CEST] <shibly> Hi, anyone there?
[18:37:48 CEST] <shibly> Anyone there?
[18:38:13 CEST] <fritsch> ask your question and check the topic of this channel ...
[18:39:01 CEST] <shibly> https://github.com/FFmpeg/FFmpeg , is it the official source code mirror of ffmpeg project?
[18:39:37 CEST] <nevcairiel> its an official mirror, yes
[18:39:52 CEST] <wm4> the official git is git://source.ffmpeg.org/ffmpeg.git
[18:40:04 CEST] <wm4> the github mirror often is behind by up to 1 hour or so
[18:40:24 CEST] <BtbN> doesn't github auto-pull?
[18:40:38 CEST] <JEEB> also why on earth is this person double-asking
[18:42:05 CEST] <shibly> If i want to be one of the maintainer of ffmpeg project, what should i have to do?
[18:42:28 CEST] <shibly> I have downloaded the source code of ffmpeg from here: https://www.ffmpeg.org/download.html
[18:42:32 CEST] <iive> start working on something and send patches.
[18:44:23 CEST] <shibly> I don't see any ticket or bug list there: https://github.com/FFmpeg/FFmpeg
[18:44:32 CEST] <JEEB> that's because that's just a mirror
[18:44:39 CEST] <JEEB> for the code
[18:44:48 CEST] <nevcairiel> shibly: start here https://www.ffmpeg.org/developer.html
[18:45:12 CEST] <JEEB> also I'm not really sure if you are going to be successful unless you have something specific in FFmpeg that you want to improve
[18:45:14 CEST] <nevcairiel> also https://www.ffmpeg.org/bugreports.html
[18:45:22 CEST] <JEEB> and yes, thanks nev for linking those
[18:45:26 CEST] <iive> btw, ffserver needs a lot of care and nobody else wants it. (wink) (wink)
[18:45:37 CEST] <JEEB> lol
[18:45:55 CEST] <JEEB> it needs to be redesigned and redone pretty much :P
[18:49:40 CEST] <wm4> could someone with access cleanup this shit https://github.com/FFmpeg/FFmpeg/pulls
[18:50:10 CEST] <BtbN> https://github.com/FFmpeg/FFmpeg/pull/201 ...what?
[18:50:34 CEST] <wm4> this happens often
[18:50:49 CEST] <nevcairiel> people try to use it as a bug report function but fail, and then this shit happens
[18:52:00 CEST] <BtbN> Hm, I wonder if it's worth it to start another discussion about whether the MIT-Licensed NVENC header makes it possible to drop nonfree with nvenc enabled.
[18:52:32 CEST] <BtbN> I think the last one was drowned out because it was also bundling the header
[18:52:37 CEST] <nevcairiel> if andreas was the only one complaining the last time, he is pretty much gone now =p
[18:53:01 CEST] <BtbN> People were primarily complaining about the bundled header
[18:53:21 CEST] <BtbN> Which I still think is a good idea, but I'll keep that for later, as it would side-track the license discussion
[21:30:28 CEST] <cone-103> ffmpeg 03Michael Niedermayer 07master:38433345e503: avcodec/utils: Assert that the number of consumed bytes in avcodec_decode_audio4() is <= the input size
[22:06:13 CEST] <ubitux> http://sprunge.us/dLRO is there a name for this algorithm?
[22:09:02 CEST] <fritsch> ubitux: https://en.wikipedia.org/wiki/Summed_area_table
[22:09:54 CEST] <ubitux> thanks, this saves me a large stupid comment
[22:11:07 CEST] <ubitux> crazy how math can make a simple thing unreadable
[22:11:30 CEST] <fritsch> that's by intention
[22:11:55 CEST] <ubitux> :(
[22:11:57 CEST] <fritsch> the problem is: formulating something 100% correct makes the notation a bit non funny
[22:12:25 CEST] <ubitux> mathematician should move to code
[22:12:29 CEST] <fritsch> also your example, seems to be a special version
[22:17:18 CEST] <fritsch> ubitux: and btw. in your image some color would have helped to understand the +a right away :-) cause the "top left coordinate" is ambigious again
[22:17:24 CEST] <fritsch> as there are many rectangles
[22:18:04 CEST] <ubitux> top-left origin of M maybe
[22:18:12 CEST] <ubitux> it's hard to put colors in ascii
[22:18:34 CEST] <fritsch> yes of M
[22:23:19 CEST] <wm4> ubitux: vt100 terminal escapes
[22:23:46 CEST] <ubitux> yeah, it will be perfect in source code
[22:23:56 CEST] <ubitux> syntax color override
[22:24:23 CEST] <ubitux> if i'm lucky it might even creates so digraph/trigraph
[22:24:30 CEST] <ubitux> to fuck things perfectly
[22:24:41 CEST] <ubitux> some*
[22:30:33 CEST] <atomnuker> color codes in C are just a step away from having HolyC's 3D models inside files having code
[22:37:29 CEST] <thardin> ffmpeg on templeos when?
[00:00:00 CEST] --- Sun Apr 24 2016


More information about the Ffmpeg-devel-irc mailing list