:: #mythtv

Daily chat history

Current users (57):

aloril, amessina, Anssi, benklop, blahdodo, brfransen, ChanServ, clever, CyberJacob, davic, eharris, ElmerFudd, frobnic, ghoti, gigem, gregl, GreyFoxx, hampton_away, Hydr0p0nX, ijc, ikevin, jafa, jarle, jheizer, jpabq, jpharvey, jya, knowledgejunkie, libsci, mad_enz, markspieth, markspieth2, MitchCapper, mkbloke, mrec_, MythBuild, MythLogBot, MythNotifyBot, nephyrin`, ooshlablu, Panic, peper03, poptix, pppingme, RamShadow, rhpot1991, rmeden, Roklobster, sdrik, ShapeShifter499, sphery_, taylorr, Tobbe5178, tris, Warped, xris-, _charly_
Friday, June 15th, 2018, 00:12 UTC
[00:12:45] gregl (gregl! has joined #mythtv
[02:07:27] SteveGoodey (SteveGoodey! has joined #mythtv
[03:01:52] SteveGoodey (SteveGoodey! has quit (Quit: Konversation terminated!)
[05:26:37] jafa (jafa!~jafa@2001:470:80ca:2000:d92b:f5ee:260c:8fce) has quit (Remote host closed the connection)
[05:29:19] jafa (jafa!~jafa@2001:470:80ca:2000:fd67:7813:974f:5966) has joined #mythtv
[06:07:52] Steve-Goodey (Steve-Goodey! has joined #mythtv
[07:14:55] peterbennett (peterbennett!~Peter_Ben@2601:183:100:3227:e990:2675:832d:3845) has joined #mythtv
[07:14:55] peterbennett (peterbennett!~Peter_Ben@mythtv/developer/peterbennett) has joined #mythtv
[07:14:55] peterbennett (peterbennett!~Peter_Ben@2601:183:100:3227:e990:2675:832d:3845) has quit (Changing host)
[07:32:44] Steve-Goodey (Steve-Goodey! has quit (Quit: Konversation terminated!)
[07:47:12] SteveGoodey (SteveGoodey! has joined #mythtv
[07:50:15] willcooke (willcooke! has joined #mythtv
[07:50:15] willcooke (willcooke! has quit (Changing host)
[07:50:15] willcooke (willcooke!~willcooke@ubuntu/member/willcooke) has joined #mythtv
[07:56:15] Merlin83b (Merlin83b! has joined #mythtv
[09:14:50] SteveGoodey (SteveGoodey! has quit (Quit: Konversation terminated!)
[10:06:39] stuarta: peterbennett: after the mediacodec stuff, were you going to look into using ffmpeg hwdec interface to the various HW decoders available?
[10:32:02] Roklobster (Roklobster! has joined #mythtv
[11:04:44] SteveGoodey (SteveGoodey! has joined #mythtv
[12:14:07] peterbennett: stuarta: I plan to look at vaapi after the mediacodec stuff
[12:17:23] peterbennett: stuarta: If you have any links to information about new vaapi features or how to implement them please pass them on.
[12:18:34] peterbennett: stuarta: I don't know about wayland, what the issues are there – I am not running wayland.
[13:25:38] dblain (dblain! has joined #mythtv
[13:25:38] dblain (dblain!dblain@mythtv/developer/dblain) has joined #mythtv
[13:25:38] dblain (dblain! has quit (Changing host)
[13:39:26] Roklobster (Roklobster! has quit (Remote host closed the connection)
[13:40:04] SteveGoodey (SteveGoodey! has quit (Quit: Konversation terminated!)
[15:07:33] stuarta: peterbennett: tbh, time would be better spend working out how to leverage the ffmpeg hwdec api, then *all* the various different hw accel supported by ffmpeg can be used
[15:17:09] peterbennett: stuarta: The existing vaapi code in MythTV has much that is strange. I would rather start from scratch.
[15:19:25] peterbennett: stuarta: While doing the mediacodec stuff I found it impossible to understand the convoluted timing code, which did not work for mediacodec, so I bypassed it entirely and that all works both for mediacodec and for s/w decoding. There seems to be lots of unnecessary complication in the code.
[15:20:10] MythNotifyBot (MythNotifyBot!~cinch@2a02:390:790f:1::1) has quit (Ping timeout: 256 seconds)
[15:20:18] stuarta (stuarta!~stuarta@mythtv/developer/stuarta) has quit (Ping timeout: 256 seconds)
[15:52:07] MythNotifyBot (MythNotifyBot! has joined #mythtv
[15:53:27] stuarta (stuarta!~stuarta@2a02:390:790f:1::1) has joined #mythtv
[15:53:27] stuarta (stuarta!~stuarta@mythtv/developer/stuarta) has joined #mythtv
[15:53:27] stuarta (stuarta!~stuarta@2a02:390:790f:1::1) has quit (Changing host)
[15:54:13] stuarta: peterbennett: sorry i got disconnected, i agree starting from scratch would be a good idea. i would do it using the ffmpeg hwdec api
[15:55:13] peterbennett: stuarta: ok
[15:55:24] stuarta: tbh moving the whole decode playback part would make our lives far simpler overall
[15:56:00] stuarta: from my reading of the api is is much more suited to properly decoding in multiple threads, and hwdec is just an extension of that
[16:00:08] peterbennett: If you have any good links with information about the h/w decoding that will be helpful.
[16:02:41] stuarta: the documentation is shit
[16:03:00] stuarta: it's basically look at how they've done the examples and try to understand wtf they are doing
[16:04:59] stuarta: AFAICT, it's create a hwdec context and feed frames to it
[16:05:53] stuarta: this might help . . . /034530.html
[16:07:26] stuarta: for bonus points. working out how to chain the deinterlacing functionality exposed by vaapi, such that the decoded frames never need to come back off the gpu
[16:10:35] ** stuarta wanders off for a bit **
[16:26:33] Steve-Goodey (Steve-Goodey!~steve@2a00:23c5:7da3:4501:9200:4eff:feb3:e38b) has joined #mythtv
[16:35:24] jpabq: peterbennett: does timestretch work on the Shield / Android TV? With !MediaCodec?
[16:36:53] peterbennett: jpabq: yes I tested it up to the maximum
[16:37:19] peterbennett: jpabq: fastest and slowest work OK
[16:37:44] jpabq: Excellent. I was wondering if that was the reason for the "convoluted code".
[16:39:22] peterbennett: jpabq: The convoluted code attempts to calculate frame timings from packet timings, however ffmpeg gives you the frame timings so you don't need to do that
[16:41:07] peterbennett: jpabq: Come to think of it, I did not test speedup with 4K video, only with 1080
[16:41:51] jpabq: I would not be surprised if ffmpeg did not handle that correclty when the mythtv code was written, but now it does.
[16:43:04] peterbennett: Yes I agree that is very likely. Also maybe ffmpeg did not give that info originally.
[16:45:38] peterbennett: The convoluted code uses a field reordered_opaque that is deprecated in favor of pkt_pts and pkt_pts is deprecated in favor of pts, so it is two levels of deprecation behind
[16:46:01] peterbennett: I imagine that pkt_pts and pts did not exist originally
[16:46:10] Merlin83b (Merlin83b! has quit (Quit: Leaving)
[16:52:57] willcooke (willcooke!~willcooke@ubuntu/member/willcooke) has quit (Quit: Do your hobbies)
[17:22:30] Roklobster (Roklobster! has joined #mythtv
[17:42:36] SteveGoodey (SteveGoodey! has joined #mythtv
[18:34:24] rmeden (rmeden! has quit (Read error: Connection reset by peer)
[19:06:24] rmeden (rmeden! has joined #mythtv
[19:44:20] gigem: peterbennett, jpabq: There is a lot of cruft left over from, as you pointed out, having to overcome limitations in external software, and also from having so many cooks in the kitchen over the years.
[19:45:39] gigem: Also, I was very interested in the timestretch performance, but was going to test it myself this weekend. I know Kodi gets kind of wonky, well, actually a lot wonky on Android, around 1.4x. I was very much hoping we could do much better.
[19:47:01] jpabq: I still have not had time to even try mythfrontend on the Shield, but I at least now own a Shield :-D
[19:47:34] jpabq: I guess I should make some time to figure out how to build myth for android...
[19:47:39] gigem: What, you bought one for markspieth, but not yourself?
[19:49:55] jpabq: I have a dream of replacing computers with the Shield, but knew I was not going to have time to do the work, so I decided to invest a little bit of money instead ;-)
[19:50:15] jpabq: Is this up to date?
[19:58:34] amessina (amessina!~amessina@unaffiliated/amessina) has joined #mythtv
[20:27:17] Steve-Goodey (Steve-Goodey!~steve@2a00:23c5:7da3:4501:9200:4eff:feb3:e38b) has quit (Quit: Konversation terminated!)
[20:39:12] jpabq (jpabq!~quassel@ has quit (Quit: - Chat comfortably. Anywhere.)
[20:39:44] gigem: jpabq: No, quite a bit is out of date. In a nutshell, the core frontend works okay to well on TV systems that are powerful enough (mainly the Shield). Slightly less powerful TV systems are probably okay for progressive content. Most TV systems should work well with peterbennett's mediacodec changes. I don't know if the gestures still work. I don't have a phone nor tablet powerful enough to run things.
[20:39:45] gigem: The phone I'm getting tomorrow probably is maybe my tablet will be with the mediacodec changes. The build procedure is best described in the packaging/android/ file. Finally, I hadn't noticed until today, but markspieth appears to be putting current, unofficial build on his website.
[20:40:49] jpabq (jpabq!~quassel@ has joined #mythtv
[20:41:30] gigem: jpabq: Check the logs. I just posted something for you without noticing you'd dropped off.
[20:42:27] peterbennett: jpabq: the instructions in git packaging/android are good
[20:42:44] peterbennett: jpabq: I updated them recently
[20:44:56] gigem: peterbennett: I'm trying to build new packages with and without your mediacodec changes, but the configure in is failing with "ERROR: opengl not found." Have you seen that before. Something must have changed on my system since I last built a package, but I'm not sure what it is.
[20:47:14] peterbennett: gigem: I have never had an error like that. Are you using latest master?
[20:47:32] gigem: Yes. I pulled this morning.
[20:48:56] peterbennett: I am using latest master for MythTV and packaging. I rebuilt libs on May 21st
[20:50:38] gigem: I built libs this morning without issue as far as I know. After the first opengl not found, I noticed my packaging was 4 commits behind. I pulled thos a little while a go, but am still getting the not found error.
[20:50:47] peterbennett: I installed Ubuntu 18.04 and rebuilt everything from scratch. makelibs took about 15 attempts to get it right
[20:52:25] peterbennett: You will need the latest packaging commit to build my patches
[20:54:35] gigem: No problem. As I said, pulled the latest and started the from scratch. Did your latest changes affect I'll keep looking as time allows. I'll find it eventually, but thought I'd ask first.
[20:55:33] peterbennett: I can supply a built package to anybody that wants it ....
[20:56:19] peterbennett: gigem: You mean makelibs? – No nothing I did new affects that
[21:01:36] gigem: peterbennett: Yes,
[21:20:28] gigem: Hmm, I'm seeing two suspcious things with "check_lib opengl GLES/gl.h glGetError -Wl,-framework,OpenGLES", which I think is the OpenGL config which should be used. First, ld complains about "-f may not be used without -shared" and collect2 complains about "ld returned 1 exit status". I think that's being ignored. Second, pkg-config complains about "Package glesv2 was not found in the pkg-config search
[21:20:30] gigem: path." I think that's the fatal part.
[21:27:11] SteveGoodey (SteveGoodey! has quit (Quit: Konversation terminated!)
[21:46:27] peterbennett: gigem: I may have changed those a while back when I was separating the ffmpeg confgure from the mythtv configure, but that was some time ago and it has been working since.
[21:54:29] gigem: peterbennett: I'm running from scratch to be absolutely sure. If that doesn't fix it, I'll suspect some subtle host change affected target configuration. I run Debian Sid and do one or two updates a week.
[22:18:43] gigem: Have any Shield owners gotten the Oreo update yet? I'm curious if it affects MythTV. I wouldn't think so, but you never know.
[22:30:09] peterbennett: I have not received the update. I saw that it had been released and then withdrawn. Not sure if it was released again after that.
[22:33:42] gigem: The release resumed this week. I'm not in the preview program so don't expect it for a week or two.
[22:34:49] peterbennett: What's with Android P? They could not find a dessert name that starts with P?
[22:35:14] gigem: peterbennett: Can you check your config.ep and tell me which OpenGL checks succeed for you. I seem to be failing all of them with "check_pkg_config GLESv2" being the which seals the overall failure.
[22:35:45] peterbennett: ok
[22:36:44] gigem: No, it just means they haven't decided on or announced the name yet. It throug me a couple of years ago when there was all the talk about Android "L".
[22:37:28] gigem: s/throug/threw/
[22:38:49] gigem: peterbennett: Also, which NDK are you using?
[22:39:05] peterbennett: GL/glx.h: fails
[22:39:36] peterbennett: windows.h: fails (obviosly)
[22:39:50] peterbennett: OpenGL/gl3.h fails
[22:39:54] gigem: I expect it to be GLES/gl.h.
[22:40:04] peterbennett: ES2/gl.h: fails
[22:40:34] peterbennett: next one is OK it looks like
[22:40:56] gigem: That should be GLES/gl.h.
[22:41:01] peterbennett: oh no sorry
[22:41:26] peterbennett: GLES/gl.h glGetError -Wl,-framework,OpenGLES gets -f may not be used without -shared
[22:42:19] peterbennett: GLESv2 GLES2/gl2.h glGetError -lGLESv2 seems ok
[22:43:03] peterbennett: EGL EGL/egl.h eglGetProcAddress -lEGL seems ok
[22:43:23] peterbennett: maybe i should hav just sent you my log
[22:43:48] gigem: Okey. That should help. Those both involve pkg-config so that's what I'll concentrate on.
[22:43:57] peterbennett: i am using android-ndk-r13b
[22:45:00] peterbennett: So your " -f may not be used without -shared" is not the problem I also see it
[22:45:02] gigem: I'll keep an eye out for your log. I haven't received it yet.
[22:46:45] gigem: No, I posted earlier I was seeing the -f without error too. I thought maybe that might be working for you. The other checks make more sense. They use the host's pkg-config and thus could be affected by my Debian updates.
[22:47:04] peterbennett: I emailed you my log
[22:47:07] gigem: I'm using ndk 15b, but I know I haven't changed that since my last successfule build.
[22:47:11] peterbennett: config.ep
[22:48:07] gigem: I got your log. Thanks. I'm about to leave work soon and won't be able to pick this back up until later this evening.
[22:48:45] peterbennett: I did add some paths to my build to make it pick up ndk things
[22:51:04] gigem: Now that I know which check is failing that shouldn't be, it should be a lot easier to figure out.
[22:51:44] peterbennett: At least I thought I did, but I cannot find where
[22:51:59] markspieth: any ndk after 13b Ive found doesnt work that well
[22:52:07] markspieth: Ive found
[22:52:33] markspieth: yes the wiki is really old. Prior to when I started on android again.
[22:53:32] markspieth: I have built things since peterbennett's ffmpeg updates. All fine from AFAIR.
[22:53:42] peterbennett: If you look at the readme in the android directory in packaging I have updated that . Including extra directories to addf to your path
[22:53:50] gigem: peterbennett: btw, have you tried any hevc recordings yet with the mediacodec support? I downloaded some of the jellyfish test recordings a while back, and not surprisingly, the hevc versios failed miserably with software decoding.
[22:55:13] markspieth: Ive had good sucess with my h96pro+. does most things. The higher the outpuyt res the worse it gets.
[22:55:22] markspieth: good value.
[22:55:57] markspieth: I havent touched things in a while though. Need to rty your mediacodec mods
[22:59:40] markspieth2 (markspieth2! has joined #mythtv
[23:00:08] gigem: markspieth: I have hopes my M8S Pro will work with well with mediacodec support. I don't need it, but it or something comparable might be good enough for my mom.
[23:00:31] gigem: peterbennett: No change with your PATH additions.
[23:01:51] markspieth: gigem: if its a S912 it should be ok even with s/w decoding. (not for 4k though). up the decoding threads to 4 and it works well. dont try too high a timestretch though.
[23:04:31] gigem: markspieth: It is an s912. It's okay for 720p, but 1080i wasnt' acceptable to me, even with one-field deinterlacing. Also, I expect a modest amount of timestretching and the s912 just coulnd't cut it.
[23:06:18] peterbennett: gigem: Yes – hevc plays ok. i only tried one
[23:06:48] gigem: peterbennett: Great.
[23:07:34] peterbennett: I think teh one I tried was 720p
[23:08:08] peterbennett: However, 4K h264 works fine where it would not with software decoding.
[23:08:30] peterbennett: I gotta go now
[23:10:49] markspieth2 (markspieth2! has quit (Ping timeout: 240 seconds)
[23:11:21] gigem: I gotta go too. Bye.
[23:12:45] markspieth2 (markspieth2!~yaaic@ has joined #mythtv
[23:37:38] amessina (amessina!~amessina@unaffiliated/amessina) has quit (Remote host closed the connection)
[23:38:06] amessina (amessina!~amessina@unaffiliated/amessina) has joined #mythtv

IRC Logs collected by BeirdoBot.
Please use the above link to report any bugs.