03:15 <ndufresne> narmstrong: I use to be able to set the primary plane with size 1x1, as a hack to make the overlay visible, but that does not seems to work anymore, is that normal ?

03:16 <ndufresne> meanwhile, I'm also trying to work on getting weston running, works if I disable composite-1 output, without GL, so far no luck with lima

03:16 <ndufresne> and gst gbm backend isn't working on that driver

03:17 <ndufresne> if someone has a success story to share, I'd be preneur

03:51 <ndufresne> ok, bumped to newer mesa, and weston starts now

03:51 <ndufresne> but all the YUV formats are gone from the dmabuf extension, so no shiny zero-copy :-(

03:53 <ndufresne> could be caused by missing dmabuf or some of it in lima, would need to look into the debugs, well, progress!

03:56 <ndufresne> ok, it's only some of it, BGRx as an example works, it's interesting it missing memory cache management, just like panfrost

06:22 <xdarklight> ndufresne: composite-1 output should be fixed by: https://cgit.freedesktop.org/drm-misc/commit/?h=drm-misc-fixes&id=43cb86799ff03e9819c07f37f72f80f8246ad7ed

06:23 <xdarklight> ndufresne: maybe narmstrong can comment on the other topics, I have no idea how that all plays together

06:35 <narmstrong> ndufresne: yep 1x1 was an hack I didn’t push, kmssink should really support overlay planes in atomic mode where all other planes are disabled

08:25 <Darkmatter66> narmstrong, I tried the nanopi-k2 steps and still the board doesn't boot the u-boot on the sdcard

08:26 <Darkmatter66> narmstrong, anything i should do while powering on the board to let it load the sdcard u-boot?

08:27 <narmstrong> nop, nothing more

08:28 <narmstrong> do you have serial hooked to see the output ?

08:28 <Darkmatter66> narmstrong, for the vendor u-boot i used the baylibre one with branch "n-amlogic-openlinux-20170606" but I don't think that makes any difference

08:29 <Darkmatter66> narmstrong, yes i have, it goes straight to the u-boot on the emmc, I will copy the logs here

08:31 <narmstrong> yeah well, you'll need to erase the uboot on emmc... emmc has the priority

08:32 <narmstrong> or you can build a HDMI dongle following https://github.com/superna9999/linux/wiki/Amlogic-HDMI-Boot-Dongle

08:32 <Darkmatter66> narmstrong, https://termbin.com/xnny

08:33 <Darkmatter66> narmstrong, damn I was hoping to test the mainline without erasing the emmc

08:35 <narmstrong> sometime there is some test points you can ground to bypass the emmc, but it's hard to find

08:36 <Darkmatter66> narmstrong, how can i erase only the u-boot from emmc ? while retaining the rest of the data on the emmc and the partition table

08:39 <Darkmatter66> narmstrong, something like this dd if=/dev/zero of=/dev/sdc bs=446 count=1 will it retain the partitions?

08:43 <narmstrong> Darkmatter66: you should save the content of the emmc before everything

09:05 <ukleinek> Darkmatter66: I would expect the partition table to be defined in the last 4*16 + 2 bytes of the first 512 byte page

14:35 <narmstrong> The partition table for eMMC is stored in the dtb not in the memory

15:04 <ndufresne> narmstrong: just one concern though, on some platforms, you must give one primary plane an FB to allow enabling any of the overlays

15:04 <ndufresne> are you saying that on amlogic, the overlay is just like primary ?

15:06 <ndufresne> Xilinx has some hack in their gst repo for that, but it does not seem to be generic enough yet

15:06 <ndufresne> There is really no library that does this right for video streaming purpose ...

15:09 <narmstrong> ndufresne: yep the overlay can be enabled alone

15:10 <ndufresne> I guess one would just try this, but in the end it will make quite a hairy code base to support both

15:10 <narmstrong> ndufresne: at least why not a fully transparent primary if overlay cannot be enabled alone ?

15:10 <ndufresne> that was the compromized I had in mind

15:10 <narmstrong> The concepts of primary and overlays are not well defined, even less their relationship

15:10 <ndufresne> right now, by default, kmssink assumes teh overlay will be over, and leave teh fbcon FB in place

15:11 <ndufresne> the reason, because that's how it worked on Intel, and was tested there ;-D

15:11 <narmstrong> Which is wrong on all other TV socs !

15:11 <ndufresne> no, it's also the case on IMX.6 and some Qualcomm SoC

15:12 <ndufresne> it's basically random

15:12 <ndufresne> well, until now, since one can read the zorder now

15:12 <ndufresne> kmssink is already pretty ancient

15:13 <ndufresne> oh, TV SoC, missed that part

15:15 <ndufresne> anyway, I'll try and figure-out something, this is one of my pet projects to get some smooth streaming on amlogic with GStreamer

15:16 <ndufresne> (while avoiding to write a complete compositor)

20:10 <xdarklight> jbrunet: thank you for the clock dt-bindings tag!

20:12 <xdarklight> jbrunet: I'll review v2 of your Libretech-PC patches during the weekend, I need a bit more time than I have left today

21:00 <xdarklight> afaerber: (I'm not sure if this is really related to your LED display series) there was a discussion about 7 segment displays like three years ago: https://lore.kernel.org/patchwork/cover/743571/ - maybe you're interested in checking the comments in there. if it's unrelated then sorry for the noise

23:41 <afaerber> xdarklight, thanks for that pointer, let's see

23:42 <afaerber> I am contributing this to the leds subsystem; the main thing that's missing there is some batch-update API for more than one LED at a time