[Ffmpeg-devel-irc] ffmpeg-devel.log.20161207

[burek]

[00:33:39 CET] <cone-485> ffmpeg 03James Darnley 07master:f33714a6948f: avcodec/h264: clean up and expand x86 function definitions
[00:33:40 CET] <cone-485> ffmpeg 03James Darnley 07master:58ca2ef62e7c: whitespace changes after last commit
[00:33:41 CET] <cone-485> ffmpeg 03James Darnley 07master:add21d0bb3f5: avcodec/h264: mmx2, sse2, avx 10-bit h chroma deblock/loop filter
[00:33:42 CET] <cone-485> ffmpeg 03James Darnley 07master:728651df06f4: avcodec/h264: mmx2, sse2, avx 10-bit 4:2:2 h chroma deblock/loop filter
[00:42:24 CET] <rcombs> atomnuker: decoding https://www2.iis.fraunhofer.de/AAC/7.1auditionOutLeader_v2_rtb.mp4 spews errors/warnings; could you confirm if they're valid?
[00:42:37 CET] <rcombs> (if they are, maybe they should be printed just once?)
[00:44:49 CET] <cone-485> ffmpeg 03Steven Liu 07master:6b95da9af91c: avformat/hlsenc: save the EXT-X-DISCONTINUITY from old list
[01:20:37 CET] <peloverde> rcombs: The set of errors I'm seeing look suspicious but it's been a while sine I've worked on the code
[01:20:54 CET] <atomnuker> rcombs: I fixed that for 5.1 such that it only gets printed once (IIRC)
[01:21:28 CET] <rcombs> atomnuker: I'm not missing any relevant commits here
[01:21:59 CET] <rcombs> (I'm behind by one commit in lavc/aacdec*.c, but it's just a change to the sample-request lines)
[01:22:18 CET] <rcombs> [aac @ 0x7f9c5382e200] Invalid bitstream - cannot apply SBR to element type 6
[01:22:18 CET] <rcombs> [aac @ 0x7f9c5382e200] element type mismatch 0 != 6
[01:22:18 CET] <rcombs>     Last message repeated 22 times
[01:22:22 CET] <rcombs> ^ I get a lot of this
[01:22:36 CET] <rcombs> which is somewhat different from the 5.1 lines, iirc
[01:23:25 CET] <peloverde> I think there is a bug surrounding elem_type_prev with multiple FIL elements
[02:11:01 CET] <peloverde> rcombs: I sent a patch to the list to fix an actual bug that I think the stream exposes
[02:12:20 CET] <rcombs> oh, heh
[02:17:10 CET] <peloverde> You "cannot apply SBR to element type 6" because element type 6 isn't audio channel(s)
[02:57:02 CET] <KGB> [13FFV1] 15michaelni pushed 1 new commit to 06master: 02https://git.io/v1E3S
[02:57:02 CET] <KGB> 13FFV1/06master 14809967a 15dericed: convert code-in-tables to tables-in-codeblocks...
[08:47:47 CET] <root> quit
[09:08:08 CET] <ElAngelo> /etc/systemd/network/MyDhcp.network
[10:18:16 CET] <superware> I need to embed "user data" per frame in h264, is this something ffmpeg (libavcodec) can do?
[10:23:43 CET] <superware> (or "unregistered data"...)
[17:24:23 CET] <cone-479> ffmpeg 03Michael Niedermayer 07master:c188f358aaee: avcodec/wavpack: Treat the first block coding too many channels as an error
[20:50:33 CET] <cone-479> ffmpeg 03Vittorio Giovara 07master:8f58ecc344a9: lavu: Add AVSphericalMapping type and frame side data
[20:50:34 CET] <cone-479> ffmpeg 03Vittorio Giovara 07master:e7a6f8c972a0: lavc: Add spherical packet side data API
[20:50:35 CET] <cone-479> ffmpeg 03Vittorio Giovara 07master:4dcdecf4719a: mov: Export spherical information
[20:50:36 CET] <cone-479> ffmpeg 03Vittorio Giovara 07master:0c0da45f0fc0: fate: Add a monoscopic spherical mov test
[20:55:20 CET] <durandal_1707> jamrial: there is user on #ffmpeg having problems compiling ffmpeg after your yasm
[20:56:16 CET] <J_Darnley> Do you mean me rather than him?
[20:56:50 CET] <durandal_1707> probably, sorry!
[20:57:25 CET] <J_Darnley> Has he given you a pastebin of the error or shall I go ask myself?
[20:59:06 CET] <durandal_1707> he is using mingw with MSVC, I told him to open bug report
[21:00:04 CET] <drv> http://pastie.org/private/wptu9nj3gra8pl7xjb1ow
[21:04:36 CET] <J_Darnley> Oh cock.  It might be the failure reported here: http://fate.ffmpeg.org/log.cgi?time=20161207200321&log=compile&slot=x86_32-msvc12-windows-native
[21:07:05 CET] <durandal_1707> J_Darnley: yup that is same assert he reported
[21:08:39 CET] <J_Darnley> Bloody msvc users!
[21:09:34 CET] <durandal_1707> heh, just drop support for them  :)
[21:35:18 CET] <J_Darnley> Can someone with 32-bit msvc test a patch on fate?
[21:35:39 CET] <J_Darnley> nevcairiel: you own the machine I was looking at on fate, can you?
[21:40:36 CET] <J_Darnley> http://pastebin.com/NhNEmnQ2
[21:41:51 CET] <BBB> J_Darnley: you can change 2*mmsize into -2*mmsize
[21:42:00 CET] <BBB> J_Darnley: since I dont think you use r%dm anywhere in your assembly
[21:42:14 CET] <BBB> J_Darnley: so you dont need the stack register at all (its just to keep access to r%dm)
[21:42:37 CET] <BBB> oh wait you use alpha_m, nevermind
[21:42:44 CET] <J_Darnley> I do *now*
[21:43:16 CET] <BBB> oh right, at least deblock_h_chroma422_10 didnt before
[21:43:22 CET] <BBB> change stack mem to a negative number
[21:43:23 CET] <BBB> that fixes it
[21:43:34 CET] <BBB> and then you should be fine
[21:43:50 CET] <BBB> (this is documented somewhere obscure but nobody knows about it)
[21:45:27 CET] <J_Darnley> That only prevents use of the original stack, right?  Because I do use it for storing mm regs.
[21:45:37 CET] <BBB> yes, argument stack only
[21:46:39 CET] <BBB> otherwise whats the point of reserving stack memory if you cant use it ;)
[21:47:15 CET] <J_Darnley> Fucking yasm!  -2 is a totally fine number!
[21:48:29 CET] <BBB> haha
[21:48:32 CET] <BBB> 0 - 2 * mmsize
[21:48:40 CET] <BBB> dont ask me, yasm is extremely obscure sometimes
[21:49:24 CET] <J_Darnley> I wonder if nasm has that same bug.
[21:49:29 CET] Action: J_Darnley reconfigures
[21:50:12 CET] <BBB> dunno
[21:50:17 CET] <BBB> but we want to support both anyway right?
[21:50:56 CET] <J_Darnley> Maybe, maybe not if yasm doesn't get updates for new instructions.
[21:51:14 CET] <BBB> for now, we support yasm
[21:51:36 CET] <BBB> not telling you what to use personally, just telling you it has to work w/ yasm for users ;)
[21:51:47 CET] <J_Darnley> True.
[21:52:07 CET] <J_Darnley> hONecA5ss
[21:52:18 CET] <J_Darnley> Ooops, pretend you guys didn't see that!
[21:54:38 CET] <Shiz> hONecA5ss
[21:55:51 CET] <J_Darnley> nasm does *not* have that bug.  I'm sold
[22:02:58 CET] <kierank> J_Darnley: don't worry it appeared as ******* for me
[22:03:03 CET] <kierank> hunter2
[22:03:19 CET] <J_Darnley> Ha.  OLD!
[22:05:53 CET] <JEEB> :D
[22:09:05 CET] <durandal_1707> is that Facebook pass?
[22:09:42 CET] <J_Darnley> No.  My user password on my "dev server"
[22:10:02 CET] <J_Darnley> Well, part of it anyway.  I alt-tabbed too soon
[22:10:28 CET] <jamrial> J_Darnley: you could have just pretended it was part of a pastebin link you already deleted :p
[22:11:11 CET] <J_Darnley> I could but that isn't very fun
[22:11:13 CET] <BBB> J_Darnley: patch is ok ;)
[22:27:16 CET] <atomnuker> BBB: that doesn't mean we should break the build for anyone using nasm
[22:27:34 CET] <BBB> I wasnt intending to do that?
[22:27:44 CET] <BBB> I think I was suggesting to unbreak yasm
[22:28:19 CET] <atomnuker> yeah, it just that you said "we support yasm" as if it's the only thing we support, nvm
[22:29:15 CET] <BBB> np, sorry for the misunderstanding
[22:35:36 CET] <Zeranoe> Well, my hosting provider did it again. Someone is scripting a bulk download, and it's crushing my bandwidth
[22:37:49 CET] <Compn> :(
[22:38:23 CET] <Zeranoe> I wish I knew who/what it was
[22:38:50 CET] <Compn> Zeranoe : i wonder if you ask a university, say UofM for some free hosting?
[22:39:09 CET] <Compn> they have internet2 setup too i think. plenty to go around
[22:39:37 CET] <Zeranoe> How ironic, I'm trying to get into that university
[22:39:47 CET] <Compn> http://its.umich.edu/
[22:39:49 CET] <Compn> haha
[22:39:55 CET] <Compn> well maybe its a good way to get in
[22:40:07 CET] <Compn> get a good contact at the it dept...
[22:40:15 CET] <Compn> they give a good word for you to get in
[22:40:28 CET] <Zeranoe> I'm not going to CS, but it would be if I was
[22:40:49 CET] <Compn> they can put in a good word for you, even if you want a liberal arts degree :P
[22:41:05 CET] <Zeranoe> Probably
[22:41:17 CET] <Compn> cant hurt to make a few calls no doubt
[22:42:54 CET] <Zeranoe> Anyone good at investigating or know of any software that would specifically use ffmpeg-20160712-df1dc52-win32-static.zip ? Whoever this is was crushing that version until I removed it, and they switched to 3.2 last night.
[22:43:19 CET] <Compn> you arent getting any referrer info ? useragent info ?
[22:44:01 CET] <Compn> how many downloads per day ? this might help us eliminate some smaller software releases...
[22:44:11 CET] <Zeranoe> I haven't looked but this is 100% scripted behavior
[22:44:16 CET] <Compn> yes
[22:44:25 CET] <Compn> but if you tell use the useragent is NSIS installer or something else
[22:44:33 CET] <Compn> we can at least make some informed guesses.....
[22:44:42 CET] <Compn> and eliminate some other players.
[22:45:19 CET] <Zeranoe> In November it was hit 173,896 times
[22:45:46 CET] <Compn> ok, then that probably eliminates something like kodi for windows 
[22:46:09 CET] <nevcairiel> big projects also ship their own customized builds
[22:46:42 CET] Action: Compn wants useragent infos
[22:47:00 CET] <Zeranoe> Looking now
[22:48:12 CET] <jamrial> Zeranoe: implement a captcha. it will take like one day for whoever is shipping a script that downloads binaries from your website to ask you why it's not working anymore
[22:48:32 CET] <Compn> jamrial : yeah, Zeranoe discussed this before. captchas suck :P
[22:48:41 CET] <Zeranoe> "Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"
[22:48:46 CET] <Compn> mmmmmm
[22:48:52 CET] <jamrial> china then?
[22:48:56 CET] <Compn> baidu spider
[22:49:01 CET] <Compn> of course china
[22:49:08 CET] Action: Compn shakes fist at china
[22:49:28 CET] <Compn> Zeranoe : your robots.txt any good ?
[22:49:44 CET] <Zeranoe> Another "Mozilla/4.0 (compatible; MSIE 8.0; Windows NT 6.1; Trident/4.0))"
[22:50:06 CET] <Compn> http://help.baidu.com/question?prod_en=master&class=Baiduspider
[22:50:29 CET] <nevcairiel> IE useragents is something people like to use for their home-made downloaders to avoid blockage =p
[22:50:37 CET] <Zeranoe> Ok, this looks like the one "Mozilla/5.0 (Windows NT 6.1; WOW64; rv:15.0) Gecko/20100101 Firefox/15.0"
[22:50:44 CET] <JEEB> &32
[22:50:59 CET] <Compn> yea thats just a generic one
[22:51:00 CET] <Compn> hmm
[22:51:06 CET] <nevcairiel> firefox 15 has to be a script, so old :D
[22:52:11 CET] <Zeranoe> One of the high traffic IPs is from Ecuador 
[22:52:23 CET] <wm4> so, ddos?
[22:52:48 CET] <nevcairiel> its not that high traffic to bring the server down, just to annoy the hoster
[22:52:57 CET] <Zeranoe> It's not consistent though 
[22:52:57 CET] <nevcairiel> (and in turn bring the server down =P)
[22:53:14 CET] <Compn> what ip in ecuador
[22:53:27 CET] <Zeranoe> Hm two IPs from Ecudaor now, crushing it
[22:53:45 CET] <Compn> they just download the same ffmpeg binary over and over again ?
[22:53:50 CET] <Zeranoe> 181.112.109.63 and 190.12.51.144
[22:53:57 CET] <Zeranoe> Yup
[22:54:58 CET] <Zeranoe> That same user agent for both, and lots of requests at the same time, but not for very long
[22:56:14 CET] <Compn> first one looks like an isp range
[22:56:49 CET] <Compn> second same thing
[22:57:21 CET] <bencoh> why would anyone keep downloading the same version anyway?
[22:58:07 CET] <Compn> if it were multiple ips, it would be users of some software that upgraded
[22:58:31 CET] <Compn> but if there are one or two ips downloading the same ffmpeg, that makes no sense to me. unless its some kind of proxy
[22:58:57 CET] <bencoh> or university / campus / whatever behind a NAT, but ...
[22:59:41 CET] <Compn> Zeranoe : easiest solution would just be to limit speed to the one or two ip addy that take the bandwidth heh
[22:59:48 CET] <jamrial> universities have their own ip ranges, usually
[22:59:59 CET] <nevcairiel> maybe not in ecuador
[23:00:07 CET] <Compn> ipv4 is running out :P
[23:00:23 CET] <bencoh> yeah I suspect ecuador joined the party a tad late
[23:00:29 CET] <nevcairiel> could be a company that has some tool on all their internal pcs that randomly all start to update
[23:00:31 CET] <cone-479> ffmpeg 03James Darnley 07master:acdd2d805d23: avcodec/h264: resolve assert being triggered when stack is not aligned
[23:00:45 CET] <Compn> but i think he means univ have their own internet hosts... they dont usually use 3rd party companies
[23:01:11 CET] <Compn> e.g. .edu domain
[23:01:19 CET] <nevcairiel> edu is really only a US thing
[23:02:02 CET] <jamrial> edu.$ccTDL is pretty common
[23:02:10 CET] <Compn> yeah edu.blah 
[23:02:22 CET] <J_Darnley> The UK decided to be special there.
[23:02:27 CET] <J_Darnley> .ac.uk
[23:02:37 CET] <nevcairiel> academic?
[23:02:49 CET] <J_Darnley> Maybe, I'd have to look it up though.
[23:03:10 CET] <Compn> plenty of other countries using .edu though. https://www.tilburguniversity.edu/
[23:03:44 CET] <nevcairiel> its more like individual universities doing so
[23:04:00 CET] <Zeranoe> Hm, I'm seeing IPs all over the board, Indonesia now
[23:04:51 CET] <Compn> is there anyway to ... uh... collect them and organize them. maybe find some pattern
[23:04:54 CET] <Zeranoe> And Cameroon. Nice, didn't even know they had internet there
[23:05:07 CET] <nevcairiel> internet is everywhere these days
[23:05:09 CET] <Compn> or give us access to the awstats for your site :P
[23:05:47 CET] <Zeranoe> They all seem to be using Firefox, with that generic string, but the version is varying.
[23:06:15 CET] <Compn> ok, i was just hoping there would be a clue , maybe language or installer version, but i guess not.
[23:06:30 CET] <Compn> is there a pattern, can you give us the exact day that the large number of downloads happens?
[23:06:39 CET] <Compn> maybe we can correlate that with the release of some video editing software ?
[23:07:07 CET] <Compn> or video game or ... this is not going to be easy
[23:07:29 CET] <Compn> why not put up a message on your website "hey, whoever is direct linking to these, please stop or identify yourself' ?
[23:08:08 CET] <Zeranoe> I was thinking about that, but they went a month before updating from the build I knew they were hitting before, so....
[23:54:01 CET] <Zeranoe> Here is a log of all requests to the suspect archive in Oct. http://zeranoe.com/shared/OSG3l8E2rvfWdhCS/traffic.zip I put it in a zip because the log is >300MB uncompressed.
[23:55:06 CET] <Zeranoe> If anyone would like to help make sense of it I would be grateful
[00:00:00 CET] --- Thu Dec  8 2016