| 09:33 | mk01_ | jnettlet: yes, the UHS I plan to discuss too - I have seen in IRC log you have reported that (and on 4.1.y it is failing to power up with -110 error). so was looking at log diffs between 3.14 and 4.1.y and added some additional tweaks I found (other commit where on the otherside already part of 4.1.y vanilla and backported to 3.14 from upstream). BUT |
| 09:33 | mk01_ | jnettlet: i still can't get it to work. |
| 09:34 | wbx | jnettlet: hi. it is a while back talking to you. are you still hacking solidrun devices? |
| 09:35 | mk01_ | I have still to try to revert the change of power_up sequence itself as for now this looks to me as the only difference in mine test tree - but those two commits were written/reviewed by RMK itself so have a little doubt it will help, but will try as I get exactly same SD card as is rported as no working |
| 10:17 | jnettlet | mk01, yeah i have a patch to fix 4.1 will send it over when i get home. |
| 10:18 | jnettlet | wbx, yep still hacking :-) |
| 10:18 | mk01_ | jnettlet: many thank for that in advance |
| 10:18 | jnettlet | 4.1 with my one liner was very solid. |
| 10:30 | wbx | jnettlet: do you know anything about kodi hdmi audio problems. crackling sound, sometimes after 3-4 minutes |
| 10:30 | wbx | jnettlet: i have no idea why. |
| 10:30 | wbx | jnettlet: analog is fine. hb1 |
| 10:36 | jnettlet | wbx, i am at the airport waiting for baggage right now. Can you post dmesg and kodi logs in pastebin and i will chwck later? |
| 10:38 | wbx | jnettlet: yeah. |
| 10:48 | wbx | jnettlet: kodi: |
| 10:51 | wbx | jnettlet: dmesg |
| 11:31 | jnettlet | wbx, are you using 50hz refresh rates by any chance? |
| 11:59 | wbx | jnettlet: where can I set this up? |
| 11:59 | wbx | it is a LG TV |
| 12:00 | wbx | but it is only audio which is crackling, video is fine all the time. using a network stream |
| 13:12 | jnettlet | wbx, ah it is an lg television. I was also having problems down in Barcelona with an lg television. I may need you to be my tester for this bug. |
| 15:49 | jnettlet | damn. mk01_, wbx, just got everything hooked up and sound works great on my home monitor over hdmi. I haven't tested my Sony TV / Soundbar yet but it was also working fine there as well. |
| 15:49 | jnettlet | It looks like this is an LG issue. wbx, I would have thought you were doing something completely wrong if I had not just had the same problem at the hotel in Barcelona. |
| 15:50 | jnettlet | I have had some timing problems with LG monitors before...I will need to dig through some of my old history to see if I can nail something down. |
| 15:51 | jnettlet | then I will need wbx to test because I don't have anything LG at home. |
| 15:53 | mk01_ | with LG's is everything broken by design. it is over 3 or four generations of TV where CEC is not much usable. |
| 15:53 | jnettlet | mk01_, yeah I know. So I am having no problems changing resolutions at all...except if a video is playing |
| 15:54 | mk01_ | YES |
| 15:54 | mk01_ | I have patch |
| 15:54 | mk01_ | for that |
| 15:54 | mk01_ | today will PR |
| 15:54 | jnettlet | that seems to send Kodi into a fit. Graphics and fbdev seem solid |
| 15:54 | jnettlet | some resolutions seem to have broken modes, so I will need to curate that list again. |
| 15:54 | wbx | jnettlet: cec works, but hdmi is broken. |
| 15:55 | jnettlet | wbx, just audio right? video works? |
| 15:55 | wbx | jnettlet: right |
| 15:56 | jnettlet | that was my problem in the hotel. |
| 15:56 | wbx | jnettlet: if you have anything to test, sent via mail |
| 15:56 | jnettlet | wbx, but it was only with Kodi. |
| 15:57 | mk01_ | jnettlet: did you update pcm-hdmi between the previous 3.14 and the new -sr ? |
| 15:57 | jnettlet | using speaker-test, aplay, or gstreamer-1.0 I had no problem with audio out to the LG |
| 15:57 | wbx | jnettlet: exactly. |
| 15:58 | jnettlet | mk01_, only if -lts updated it, or fsl did |
| 15:58 | mk01_ | yes this is what I mean |
| 15:58 | jnettlet | wbx, okay so we are seeing the same problem. good to know. |
| 15:59 | mk01_ | FSL introduced a setting which alsa reads as only format supported. mine samsung had no sound too. |
| 15:59 | mk01_ | wait, will check the source |
| 15:59 | jnettlet | no sound, or sound only plays as static? |
| 16:08 | jnettlet | mk01_, so the only difference with the drivers I am using and the ones you are, are wayland libs, so for fbdev support the libraries are identical |
| 16:08 | mk01_ | jnettlet: |
| 16:08 | jnettlet | So I am interested why your display switching is crashing and mine isn't |
| 16:09 | jnettlet | mk01_, so just removing that fixes sound for you? |
| 16:09 | mk01_ | yes |
| 16:09 | jnettlet | okay I will test that here. |
| 16:10 | mk01_ | because it forces ALSA think that it is 24bit packed in 32bit |
| 16:10 | jnettlet | then I will try and build an OpenElec image for testing. wbx...what platform are yo uon? |
| 16:10 | mk01_ | and XBMC is choosing wrong format |
| 16:10 | jnettlet | yeah, I noticed that. |
| 16:10 | wbx | jnettlet: self compiled. |
| 16:10 | jnettlet | wasn't sure it was coming from. |
| 16:11 | jnettlet | wbx, can you remove the line that mk01_, pointed out above and test? |
| 16:13 | mk01_ | jnettlet: this is the screen reconfiguration - |
| 16:13 | mk01_ | if you would be recompiling xbmc |
| 16:13 | mk01_ | anyhow I will send it later today |
| 16:14 | jnettlet | hmmm, that hdmi patch is from ti |
| 16:14 | mk01_ | jnettlet: and if you are at come already, can you send me the uhs oneliner please ? |
| 16:14 | jnettlet | mk01_, yeah hold on |
| 16:16 | jnettlet | mk01_, this fixed the problem for me with all my cards. |
| 16:16 | jnettlet | I have a couple of transcend, a couple of versions of sandisk, and a few samsungs |
| 16:16 | mk01_ | that one I cherrypicked ! |
| 16:16 | jnettlet | it didn't fix it? |
| 16:16 | mk01_ | and the user is stil telling me it doesn't power up |
| 16:16 | jnettlet | what version HB do you have? |
| 16:17 | mk01_ | he is on Cubox. is story different there ? |
| 16:17 | jnettlet | oh yeah |
| 16:17 | jnettlet | Cubox doesn't have the proper circuitry |
| 16:17 | jnettlet | next iteration will |
| 16:17 | jnettlet | I need to patch the device-tree to force it to no-1-8-v |
| 16:18 | jnettlet | The Cubox-i supports UHS cards, but only at SDR speeds, the HB actually runs at UHS speeds. |
| 16:19 | jnettlet | It was actually a lot more work than expected to get UHS to work reliably. |
| 16:21 | mk01_ | I remember that 'no-1-8-v;' setting vanished out after some ?rmk's? patches (on the way to 4.1). so all it is about this not being supported in driver/mmc tree? |
| 16:22 | mk01_ | still, I will get that non working SD later this day so definitely confirm the working HB |
| 16:23 | jnettlet | mk01_, yeah I need to get this sorted out in upstream. |
| 16:23 | mk01_ | jnettlet: how can I get any logs from the res switching ??? beside the EGL layer returned error and hard freeze after 2s there is nothing visible |
| 16:23 | mk01_ | (nothing I mean no kernel panic, etc) |
| 16:23 | jnettlet | mk01_, you will need to edit the drivers/gpu/galcore/config and enable debugging there. |
| 16:24 | mk01_ | it is very very hard freeze ;) |
| 16:24 | jnettlet | hopefully we will see something useful on the console |
| 16:24 | wbx | jnettlet: the .sig_bits should be removed? will test |
| 16:25 | jnettlet | wbx, according to wbx yes :) I just reverted the patch that added it and will test here. If it works for you, mk01_, and myself then I will push the change |
| 16:29 | jnettlet | mk01_, also if you are messing around. Please install libjemalloc and preload that before starting kodi. |
| 16:29 | mk01_ | OK |
| 16:32 | jnettlet | I think it tends to make the transitions smoother when the video is playing in the background |
| 16:32 | mk01_ | jnettlet: is it safe into /etc/ld.so.preload or better not? |
| 16:33 | jnettlet | mk01_, I think best not. I generally add it to the top of the /usr/bin/kodi script |
| 16:33 | mk01_ | yes ok, np |
| 16:34 | mk01_ | btw: have you tested the new xbmcs decoder on H1? I have the feeling that the old was better on it. but it is just feeling - wanted to get it confirmed. |
| 16:35 | jnettlet | mk01_, I have just been working on getting it all stable :) |
| 16:35 | mk01_ | ah ok |
| 16:35 | jnettlet | I will start performance tweaking soon |
| 16:37 | jnettlet | I have had good success running glslopt on the shaders though |
| 16:37 | mk01_ | this was anyhow solely to xbmcs change in the 'codecimx'. me feeling was from running it on the same kernel, libs etc. just new xbmc. |
| 16:39 | mk01_ | with LD_PRELOAD=[thelibyouposted] kodi doesn't start ... looking whty |
| 16:40 | jnettlet | mk01_, are you thinking 14 vs 15? |
| 16:41 | mk01_ | yes 14 vs 15 |
| 16:42 | mk01_ | strange. started from within gdb (which has LD_PRELOAD in session) is fine |
| 16:42 | jnettlet | odd |
| 16:42 | mk01_ | if gdb has preloaded lib, a process started from within has too, right ? |
| 16:42 | jnettlet | yes |
| 16:42 | jnettlet | if you break on a malloc and then backtrace it you should see it going through jemalloc |
| 16:43 | mk01_ | hm, indeed |
| 16:43 | mk01_ | then I just messed up startup script probably |
| 16:44 | jnettlet | do you want me to tar up my directory of optimized shaders as well? |
| 16:45 | jnettlet | but jemalloc is nice because it helps with memory fragmentation as well |
| 16:45 | jnettlet | that helps to prolong some *ahem* CMA issues that are still trying to be solved better. |
| 17:02 | mk01_ | ok. so it should be handy specially (but not only) on HB1. I see CMA trying hard a lot of times |
| 17:02 | mk01_ | ((when trying to find a block)) |
| 17:03 | mk01_ | jnettlet: do you want me to tar up my directory of optimized shaders as well? |
| 17:03 | mk01_ | yes |
| 17:09 | jnettlet | mk01_, okay need to run to the grocery first. I will adopt the ZONE_CMA patchset as well and test that on the HB1 to see how it helps. Then numbers they posted seem impressive |