09:50 <lynxeye> austriancoder: the message stuck in moderation due to "Too many recipients to the message" on the etnaviv ML is pretty annoying. Is there any way to turn this off?

09:59 <austriancoder> lynxeye: lets see..

10:02 <schurig> okay, I found the error of my "libGL error: failed to authenticate magic". So here is the report for the next googler ...

10:03 <schurig> when experimenting with the X11 etnaviv/modesetting, X11 didn't start at all, because part of it did access /dev/dri/card0 instead of /dev/dri/card1. I tried to fix this with "kmsdev".. but somehow this didn't work out

10:04 <schurig> all I get was a starting X11 server, where clients like urxvt could connect, but a totally black screen

10:04 <austriancoder> lynxeye: you should never see this message again

10:04 <schurig> when I, by hand, swapped card0 and card1 device nodes, then at least X11 started (without any xorg.conf file) in framebuffer mode

10:04 <lynxeye> austriancoder: thanks!

10:05 <schurig> so I wrote systemd service that swapped it for me ... which in then forgot. And the Section / Driver "Armada" / EndSection didn't like it. Somehow X11 started, and urxvt was there, but I got this error message when I started glxgears

10:08 <schurig> Question: with Section / Driver "armada" / Endsection in xorg.conf it is normal that I only get 2D acceleration, e.g. for bitblitting and the odd program that still uses ancient X protocols ... but no 3D acceleration, e.g. it is normal that glxinfo shows me "llvmpipe". Correct ???

10:09 <lynxeye> schurig: No, at least if you built the armada driver with dri support, you should also get 3D accel.

10:09 <lynxeye> TBH though I haven't tested this in a long time

10:10 <schurig> hmm, at configure time it said DRI2 enabled, DRI3 disabled

10:10 <schurig> should I compile with --enable-dri3 ?

10:14 <lynxeye> schurig: dri2 should be fine, so something other is broken

10:16 <schurig> relevant configure output: https://pastebin.com/7A71viVV

10:29 <schurig> hmm, in the Xorg.log I found this:

10:29 <schurig> (II) Initializing extension GLX

10:29 <schurig> AIGLX: Screen 0 is not DRI2 capable

10:31 <schurig> yet kmscube, which also uses dri, works nicely

10:33 <lynxeye> schurig: kmscube does not use the X11 dri protocol

10:36 <schurig> the Xserver itself, at this time of startup, also doesn't use the X11 dri **protocol**. But it uses DRI.

10:36 <schurig> and kmscube also uses DRI. E.g. when I run "strace -oa -efile kmscube" and later "grep dri a" I see lots of hits to /dev/dri/* files, to .drirc

10:50 <schurig> IMHO all of this is totally overcomplex, buggy, and underdocumented. And a bit like on Windows. When one googles, one finds lots of result, from years ago. No up-to-date readme giving examples, expected outputs, and hints on how to debug things systemtatically ---the "error" messages don't say someting that is usally, they could as well jus say "some unknown error happened, you're hosed".

10:50 <schurig> sorry, but I'm really frustrated

10:50 <schurig> I know that the devs all work on their free time on it ... but so did I for my opensource programs

10:51 <schurig> on windows, the error messages are also next to unusable, and you also hear "just update", without a way of finding the root cause ... or gaining a thorough understanding

10:55 <lynxeye> schurig: Xorg has never been a well maintained platform with etnaviv, as too few people are interested in this

10:57 <lynxeye> /dev/dri is the DRM device. kmscube is using this to do render acceleration. To get 3D accel under X11, the OpenGL driver aka Mesa has to talk to both the etnaviv DRM device for the acceleration AND the X server for presentation via the DRI2 protocol

10:58 <schurig> so it should be named /dev/drm then .... see what I mean with overcomplex? I, in my simplicity assumed that if semething is /dev/dri, then it's DRI. This is not personally meant to you, it's totally not your fault.

11:00 <lynxeye> naming is hard and sometimes names picked some decades ago does turn out to be confusing for other people...

11:02 <schurig> btw, kmscube also (successfully) ldopens libxcb-dri2.so and libxcb-dri3.so ... so there are more things that made me assume that it uses "DRI"

11:02 <schurig> so, if now the screen claims to not support DRI ... how can I dig into this?

11:32 <Marex> schurig: I totally understand your frustration, when I debug etnaviv, I am going through the same ... endless reading of the code, since there is either no documentation, or it's scattered all over the various github repos ... but at least I dont have to deal with X11

11:33 <schurig> i'm now starting to put the equivalent of printf's into the code :-/