**** BEGIN LOGGING AT Tue Jul 02 02:59:56 2019 Jul 02 04:05:37 i can't wait for the buster builds! Jul 02 04:07:10 i'll at long last have netatalk 3 :D Jul 02 04:16:30 there are testing builds for buster, or you can use apt to upgrade (but I'd suggest only doing that with a console image) Jul 02 04:17:03 i thought about that but have a couple questions Jul 02 04:17:11 fire away Jul 02 04:20:12 1) if i upgrade to testing now, in a little under a week when testing is scheduled to become the new stable, is it just a matter of updating some text in the apt source files and doing an `apt upgrade` to make it as if i'd waited for the release or will there be residual differences? Jul 02 04:20:54 if you configure your sources to "buster" then you won't even notice when buster transitions from testing to stable Jul 02 04:21:53 ok, cool Jul 02 04:22:29 2) is there anything about the machine being a beagle (e.g. drivers) that complicates doing the dist-upgrade to testing in the first place? Jul 02 04:22:47 I've had zero trouble Jul 02 04:23:42 of course the more stuff is installed, the more opportunity there is for something to go wrong (in addition to requiring more temporary disk space), which is why I suggested only doing this with the console image (and not iot/lxqt) Jul 02 04:31:38 ok. yeah, i have the console image without much added to it yet. Jul 02 04:34:18 i just remember when i was using rcn's *ubuntu* console images, every time depmod ran during an apt upgrade i'd get errors like the following (with just the version varying) Jul 02 04:34:21 WARNING: missing /lib/modules/4.19.31-armv7-x31 Jul 02 04:34:22 Ensure all necessary drivers are built into the linux image! Jul 02 04:34:41 haven't seen that since switching to the debian images, though Jul 02 04:36:26 that sounds like something was horribly broken Jul 02 04:36:47 I've never seen that Jul 02 04:40:55 here's an example of what i got immediately after flashing one of the ubuntu console images: https://pastebin.com/855GEt9q Jul 02 04:42:08 it would do the same after flashing a newer image later Jul 02 04:42:48 sounds pretty bad Jul 02 04:42:56 led me to think there was something to the image build process that i couldn't get by just `apt upgrade` on the system Jul 02 04:43:13 no, this just sonuds like the ubuntu image is broken Jul 02 04:43:29 no idea what could be going on there Jul 02 04:44:00 good to hear :) Jul 02 04:45:17 just fyi, rcn-ee[m] (kindly see preceding 15 mins of messages) Jul 02 05:08:41 zmatt: is there any reason i should or shouldn't `upgrade` before `dist-upgrade`? Jul 02 05:14:36 shouldn't matter, doesn't hurt Jul 02 05:16:52 basically 'dist-upgrade' is just a bit more aggressive and may be required in some cases where plain 'upgrade' will leave some packages unupgraded Jul 02 06:10:55 zmatt: thanks. now i accidentally retained stretch's /etc/securetty file. once this finishes, how can i most easily fix that? Jul 02 06:12:01 a modified (by you or by rcn) version of stretch's /etc/securetty I presume, since otherwise it would just have silently replaced it Jul 02 06:12:31 did it leave a /etc/securetty.dpkg-dist ? since then you can just compare them to investigate what you want to do Jul 02 06:18:45 rcn must've modified it. how can i see what stretch's original version was? i could research the differences to see if they would logically carry over to buster (which i'm guessing they probably would). i'll have to wait til the dist-upgrade finishes to see if it's left the .dpkg-dist files behidn Jul 02 06:21:42 I know dpkg keeps the originals of config files somewhere, but I don't actually know how you're supposed to locate those Jul 02 06:33:45 I personally also use etckeeper to make it easier to keep track of what happens in /etc Jul 02 13:05:10 Hi, we have a custom board based on AM3703. We were using 4.1 and 4.4 mainline kernels so far. I've tried an upgrade to mainline 4.19 and it basically works after minor changes. But MUSB works only in PIO mode, no DMA. Jul 02 13:05:47 I've tried 4.19 mainline with our custom defconfig. I've also tried 4.19-ti with custom defconfig and with omap2plus_defconfig also. Jul 02 13:06:33 We have some patches to enable our board (our DTS, cusom touchscreen driver and other smaller stuff). Jul 02 13:07:25 The devices on MUSB are recognized (e.g. visible in lsusb), but the communication doesn't work. Jul 02 13:07:48 Do you have any pointers what should I check or test? Jul 02 13:21:01 Maybe a basic question first... Is it DMA on MUSB on AM3703 expected to work on 4.19? Jul 02 13:37:42 dabbler1: update-initramfs: Generating /boot/initrd.img-4.19.31-armv7-x31 Jul 02 13:37:42 WARNING: missing /lib/modules/4.19.31-armv7-x31 Jul 02 13:37:49 that's just noize of the "dual" kernel elinux images.. Jul 02 13:38:33 i didn't care to generate one image for the am335x and another for beagle-xm, so you see noise from teh "other" kernel installed.. Jul 02 13:39:25 tnovotny: give up on "dma" on musb, it's broken, only works for a specific set of instances.. Just don't bother.. Jul 02 13:40:14 rcn-ee[m]: ouch...thanks a lot for the information. Is there a ~big speed impact? Jul 02 13:40:49 BTW it is broken because of dma subsystem changes in Linux? Jul 02 13:41:55 forget the impact, I will measure by myself. Jul 02 13:42:09 tnovotny: speed impact? It was never designed for "real" usb devices, it only supports a mythical special usb interface that doesn't actually exist.. Jul 02 13:42:30 consider it foo-bar'ed.. Jul 02 13:43:37 ok, it is a good message in the end. Jul 02 13:44:41 And one more question: is there general advice which kernel to use for omap3? I mean mainline/ti/beagleboard/something_else? Jul 02 13:44:59 or it depends on use case? Jul 02 13:47:10 for omap3, you can start with mainline omap2plus_defconfig or multi_v7_defconfig Jul 02 13:47:44 great, this I did so far...thanks a lot again. Jul 02 13:48:37 the "beagleboard" "ti" kernel doesn't really support omap3 (or omap4) as TI's kernel fork (v4.19.x/v4.14.x) doesn't offically support it.. (aka they could break it and wouldn't care).. so for omap3 mainline is really best.. Jul 02 13:49:38 thanks for explanation Jul 02 14:05:16 zmatt: interesting, thanks for mentioning etckeeper Jul 02 14:07:25 rcn-ee[m]: oh, i see. but that only applies for ubuntu and not debian? Jul 02 14:10:12 dabbler1: nope.. Jul 02 14:11:13 Both: images referenced on: https://elinux.org/BeagleBoardDebian & https://elinux.org/BeagleBoardUbuntu do that.. Jul 02 14:12:13 other then the ros-ubuntu image, to have "ros".. we didn't do a pure "BeagleBone" "ubuntu" image.. Jul 02 14:20:24 rcn-ee[m]: oh, it applies for your non-BBB.org images, but not the ones in https://elinux.org/Beagleboard:BeagleBoneBlack_Debian#Stretch_Console_Snapshot, for instance, as there are separate ones for the two processor situations? Jul 02 14:35:01 @da Jul 02 14:58:22 dabbler1: those don't have the issue, as i have a special image for the xM Jul 02 14:58:27 m Jul 02 16:42:57 hello, is somebody how to run h264 decoder on BBBW with hardware accelertion? Jul 02 16:47:27 vadiml: what hardware ip block do you want to use for that? Jul 02 16:50:20 vadiml: you must be confusing things with some other board, the AM335x has no hardware acceleration for video Jul 02 17:00:18 zmat: i Wonder if there is non gstreamer pluging which is using NEON instruction for h264 decoding Jul 02 17:01:01 sorry, If there IS such gstreamer plugin Jul 02 17:01:21 oh you mean neon. i.e. you're looking for decent-quality software codecs Jul 02 17:03:10 yes i need to get 30 fps on 800x600 stream... Currently the best result i have with mplayer at around 15 fps Jul 02 17:07:27 oof, I don't know if there's any hope of getting 800x600 30fps H.264 decoding on the beaglebone. H.264 is just a very a cpu intensive codec Jul 02 17:09:10 vadiml: give mplayer or vlc a go also, your going to be cpu limited.. Jul 02 17:09:36 you can try recompiling software specifically for cortex-a8 + neon, that *might* squeeze out some more performance Jul 02 17:13:42 if possible, use a less cpu-intensive codec Jul 02 17:14:06 (or a board that's better suited for video playback) Jul 02 17:17:59 zmat: i can i trying to run miraclecast (miracast server) on my BBBW... It's h264 and nothing else. Vlc has problem of dropping late frames so it display couple of frames in the beginning and then nothing... Gstreamer has the same problem.. Best results are with mplayer but it does only 15fps Jul 02 17:22:01 I can't say I'm surprised. it's just not the best choice of hardware for this particular application. the am335x is an industrial processor, video playback was not a consideration in its design Jul 02 17:23:00 its graphics capabilities are basically the minimum you need for a basic touchscreen interface Jul 02 17:23:41 (they still threw in the basic 3d gpu because it's a hard requirement for android) Jul 02 17:38:35 zmatt, did you see with j7xx ti finally upgraded the 3d unit.. except anotehr sgx.. rouge gpu.. Jul 02 17:39:05 was hoping they'd jump ship.. Jul 02 17:39:26 "j7xx" ? Jul 02 17:41:35 also "finally"? the omap5 and its descendent am57xx/dra7xx already has a significant upgrade compared to the omap3 and am335x Jul 02 17:41:38 found it on meta-ti last week, it's a multi core A72 Jul 02 17:42:00 looks like keystone 3 branding.. Jul 02 17:43:01 J721E SoC, Dual Cortex-A72s in a single cluster, 3D GPU PowerVR Rogue 8XE GE8430 Jul 02 17:43:27 http://www.ti.com/lit/ug/spruil1/spruil1.pdf Jul 02 17:43:33 no surprise there given that the am654x also looks keystoneish even if not branded as such Jul 02 17:44:59 what gpu does the am654x use anyway? the datasheet and TRM just say "GPU" with some specs, but not which GPU it is (and I'm too lazy to try to match specs against powervr's offerings) Jul 02 17:46:55 13496-page TRM o.o; Jul 02 17:50:52 there's clearly a lot of similarity between the j721e (is that its codename? weird) and the am6 Jul 02 17:58:37 zmatt; After so;e disgging i see that libavcodec in BBB image is symlinked to one one in /lib/arm-linux-...../neon/vfp/libavcodec.so so it is already using NEON instructions. so it eems dead end. Thanks for the input anyway Jul 02 18:04:20 zmatt, rcn-ee[m]: both AM65x and J721e are branded as part of new K3 family. first is A53, second is A72. both have PowerVR GPU - first has SGX, latter has Rogue. Jul 02 18:04:46 I see the J721e has a new DSP too, C71xx Jul 02 18:05:36 zmatt: yep, that's a beast - first DSP with MMU Jul 02 18:41:49 three pairs of cortex-r5f cores o.O Jul 02 18:42:39 interesting that it reduces the number of cortex-A cores in favor of those Jul 02 19:30:55 ohh, nice application of PRU => https://www.youtube.com/watch?v=2qe4W_USweE Jul 03 00:59:35 zmatt rope memory emulator... wow... Jul 03 02:16:02 I am learning! Jul 03 02:48:12 LEARNING! Jul 03 02:51:37 Sorry. Just learning again. **** ENDING LOGGING AT Wed Jul 03 03:00:28 2019