IRC log of #cubox of Sun 09 Aug 2015. All times are in CEST < Back to index

08:45 cbxbiker61 jnettlet, i just installed your openelec image, the initial setup flow is real nice
08:45 jnettlet SPDIF isn't working, and there are a few rendering bugs.
08:46 jnettlet I am looking into scaling the UI to see if that makes things a bit smoother. always render the UI and scale it up for 1080p displays
08:47 cbxbiker61 where is the .config for the kernel? you built from the solid run fslc kernel tree, right?
08:47 jnettlet I am using a slightly tweaked version of the defconfigs we provide.
08:48 jnettlet it is really a work in progress.
08:48 cbxbiker61 i built a kernel from the fslc tree, but for some reason my mmcblk0 is pushed back to mmcblk1, like there are two mmc drivers coming into play
08:48 jnettlet but has everything been pretty stable for you? That is what I am most looking into. The last few OE builds have not been received well by some customers
08:49 jnettlet cbxbiker61, did you update the dtbs from the kernel source?
08:49 cbxbiker61 the fslc dtbs need to be updated?
08:49 jnettlet freescale patched things which re-ordered all the mmc devices. I had to change the device-tree to fix this.
08:50 jnettlet don't even ask me why they would think about changing such a thing.
08:51 cbxbiker61 your oe image looks ok to me, although i haven't used it much
08:52 cbxbiker61 passthrough is working, although i should probably run some high def audios through it
08:52 jnettlet it is really just a jumping off point. I asked over in #openelec and they seem to have lost all interest in support iMX6 and are just focusing on the new WeTex
08:52 jnettlet Wetec
08:52 jnettlet I was hoping to have more time to fix spdif this weekend but that just wasn't in the cards.
08:53 cbxbiker61 that's sad that they're jumping ship, but since i roll my own kodi, i'll keep my stuff working as long as it makes sense
08:54 jnettlet but I think that is the last major issue. Then I need to add a few things that OE doesn't support like running irqbalance, rngd, and preloading libjemalloc into the XBMC heirarchy to help with memory fragmentation.
08:54 jnettlet what platform do you build on top of?
08:54 cbxbiker61 xilka, which is mine from the ground up
08:55 cbxbiker61 it's pretty easy to keep it moving forward, the first two years were the majority of the work
08:56 cbxbiker61 as far as kodi distros go, i haven't figured out why nobody is rolling in nfs4 support
08:59 jnettlet I have no idea. That is one of those things where I would prefer to leave it up to the distro maintainers to decide what their users need.
09:00 cbxbiker61 i've been using nfs4 for years and really hate to support the lower versions
09:01 wbx mk01_: ping
09:02 jnettlet well nfs2 should literally just be forgotten.
09:03 wbx jnettlet: i think nfs2 is still default in linux kernel, when doing nfsroot :(
09:04 jnettlet wbx, yeah because it is very simple. No locking :\
09:04 wbx cost me some hours to find it out ;)
09:04 wbx jnettlet: but nfs-utils defaults to nfsv3, so you get a mismatch in a default setup. ugly.
09:05 jnettlet fine for being simple and lightweight but is not very robust. But I believe nfsroot is readonly right?
09:05 wbx no, can be rw.
09:06 wbx i use it with rw normally to save reports from a testsuite.
09:06 jnettlet cbxbiker61, if you support SDL2 in xilka then make sure to get version 2.0.4 and build it with Vivante support
09:08 cbxbiker61 yes
09:09 cbxbiker61 another interesting aside is that Qt5 is working quite nicely on my arm boxes, i.e. rpi cuboxi
09:10 cbxbiker61 i built a nice little touchscreen frontend in qt5 that runs real nice on a framebuffer touchscreen/rpi1
09:12 cbxbiker61 it replaced a frontend that was written in pysdl, and was just too laggy to be practical
09:15 cbxbiker61 actually that was pygame
09:21 vpeter [08:52] ... I asked over in #openelec and they seem to have lost all interest in support iMX6 ... I'm still on one other imx6 board but my knowledge to actually dev something big is limited.
09:22 jnettlet vpeter, I was talking about the core developers.
09:23 cbxbiker61 the thing i hate is product churn, and one thing that open source gives us is the ability to support older, still quite capable hardware
09:23 jnettlet mk01_ is still adding patches as well
09:24 jnettlet however nobody seems interested in supporting the instability and performance degradation that exists in the recent builds both for stable and beta channels
09:24 vpeter I would say there is not much people who understand this.
09:32 jnettlet well that is why I was offering to help get things working better, but they don't seem interested.
09:32 jnettlet I will try to get a stable build that we can ship and then work from there I guess. Not really what I wnated.
09:32 jnettlet wanted
09:36 vpeter jnettlet: Only few people are interested in imx6 on openelec channel. You, (probably) sraue (who is MIA for some time already) and me.
09:36 jnettlet well I guess I will just bring the development here then.
09:38 jnettlet since all the video rendering is now done through the IPU and gpu's 2d api's really we just need to find the solution that renders the gui the fastest possible
09:40 cbxbiker61 i do notice that you image seems to keep kodi cpu utilization down quite well
09:40 cbxbiker61 your image
09:41 jnettlet well part of that is the new kernel, the other part is I have enabled dirtyregions by default.
09:41 jnettlet with the gpu fixes you can enable that without getting the strange flickering
09:42 jnettlet although there are a couple of rendering problems but I have a feeling they are gui bugs and not the rendering fault.
09:48 cbxbiker61 jnettlet "ethtool -s eth0 speed 100 duplex full" still seems to be useful to eliminate dropout with high-def audio
09:49 cbxbiker61 i'm watching buckaroo banzai, dst-hd-master, with 100Mbit, no dropouts
09:49 cbxbiker61 dts-hd-master
09:50 wbx cbxbiker61: over hdmi?
09:50 cbxbiker61 yes
09:50 wbx cbxbiker61: cubox-i or hb?
09:50 cbxbiker61 cubox-i4
10:13 jnettlet cbxbiker61, oh I know why.
10:13 jnettlet it is because OpenElec doesn't include our irqbalance daemon.
10:14 jnettlet I bet under those circumstances you can generate enough interrupt contention that it causes drop outs.
10:14 jnettlet I will be interested to see if the next test build fixes that.