05:59 <tomeu> macromorgan: that probably can be addressed in gbm-gralloc

08:44 <tomeu> macromorgan: so glodroid works with panfrost on the allwinner h6, which has a mali t720

08:45 <tomeu> see the branches in this repo: https://github.com/GloDroid/glodroid_forks

08:45 <tomeu> some of the patches they are carrying for mesa, gralloc and drm_hwcomposer might be relevant

08:45 <purringChaos> I have stumbled across a bug and have no idea who to report it to so I made a full writeup and steps to reproduce it here https://gist.github.com/purringChaos/adbb1d76d3aa91b838879b684e7bf4b5 I'm forwarding this message to everyone in the chain of where this issue may be. Keep in mind I do not have a IRC bouncer and can't use matrix so I have my

08:45 <purringChaos> contact info inside the gist. I hope this is helpful enough.

08:46 <tomeu> purringChaos: this is the place to report bugs: https://gitlab.freedesktop.org/mesa/mesa/-/issues/

08:47 <purringChaos> I can't use the gitlab as it says i do not have permission to create issues or comment on issues or i have been blacklisted, i tried there first, if someone could make a report for me that would be good

08:48 <purringChaos> Might just be this IP address who knows

08:49 <warpme_> guys: I give run of current mesa master on t720 and getting... "Cannot create platform OpenGL context, neither GLX nor EGL are enabled". It looks like master on t720 offers nothing. Is this expected WiP or regression?

09:00 <macc24> warpme_: huh? it worked fine on my machine

09:01 <warpme_> hmm - what kernel?

09:01 <macc24> what build options do you have enabled?

09:01 <icecream95> purringChaos: Are there any messages in dmesg?

09:02 <macc24> warpme_: mainline 5.8.1

09:02 <macc24> yes it is outdated

09:02 <warpme_> well - i think it is not issue of building as i have fully working 20.2.1. only change is: git hash from 20.2.1 to master and recompile

09:03 <warpme_> i'm on 5.9.1. but - let me check all again.....

09:03 <macc24> this works for me: https://bpa.st/YZIQ

09:05 <purringChaos> icecream95 none which i can see

09:06 <tomeu> purringChaos: can you report your problems entering issues in #freedesktop?

09:07 <tomeu> warpme_: I would first check with strace that the right mesa libs are being picked, and that the right file in /dev/dri/ is opened

09:19 <warpme_> ah - lack of GLX/EGL was reboot issue. I updated files on target and done reboot (warm reset). After cold reset it is better: EGL is present but screen is like this: https://paste.pics/21cbfb2a39a856437f7ce1f3429531e4 Issue seems to be in masa-drm part as vnc to machine shows correct screen. this is on today's master & 5.9.1 kernel.

09:20 <warpme_> and only t720 seems to be affected. t820/g31 works ok

13:30 <tomeu> warpme_: I would suspect display driver

13:30 <tomeu> we have a t720 in our CI, but it tests without display

13:42 <warpme_> tomeu: so it looks like bisecting will be my friend.... ough. May you hint me on which mesa source files I should look to elect "convenient" githash (i mean githash requiring not to many bisecting iterations)?

13:53 <daniels> honestly, it's not worth trying to cheat bisect

13:53 <daniels> you're almost 100% guaranteed to screw it up and cause yourself _extra_ steps

14:25 <robclark> tomeu: wait, https://github.com/GloDroid/glodroid_forks is single git tree containing mesa and multiple kernel forks, etc, on different branches? That is a rather special way to use git.. (also I think we should discourage various androids from using mesa forks.. anholt/krh have been enabling android builds in gitlab CI and landing various things so that AOSP wouldn't have to cause problems with broken downstream

14:25 <robclark> hacks)

14:25 <daniels> agree!

14:27 <robclark> I think we can get to point where you can just pull android binaries from gitlab-ci artifacts

14:34 <krh> we'll be a blob eventually!

14:51 <macromorgan> any thoughts on what I should look at in gbm_gralloc? Also I'll see what patches I need to mess with; for now the only change I've done to Mesa upstream master is to change the path to python to use the system one (since AOSP doesn't have mako in its branch)

14:51 <macromorgan> but I've tried using all the glodroid specific forks and stuff for weeks with no avail

15:25 <alyssa> krh: "we'll be a blob eventually! " lolol

15:26 <alyssa> I can reproduce purringchaos's bug

15:36 <tomeu> robclark: I found out about a minute before I wrote that :)

17:44 <warpme_> re: t720 screen issue reported by me today at 11:19AM - here is what bisect tells to me:

17:44 <warpme_> https://www.irccloud.com/pastebin/xGubXW5U/

17:46 <alyssa> warpme_: ...what?

17:46 <alyssa> that shouldn't even touch panfrost

17:48 <tomeu> bisection is hard :)

17:50 <alyssa> alas

17:57 <daniels> Bisecting: 25 revisions left to test after this (roughly 5 steps)

17:57 <daniels> [piotro@minimyth-armv8 git-src]$ git bisect good

17:57 <daniels> [f74186baa5efd1147916b39a819fee9e427c8ecd] panfrost: XMLify Midgard samplers

17:57 <daniels> [piotro@minimyth-armv8 git-src]$ git bisect bad (black screen)

17:57 <daniels> that seems like a pretty obvious candidate

18:33 <tomeu> thought it wasn't black screen, but rather https://paste.pics/21cbfb2a39a856437f7ce1f3429531e4 ?

18:33 <daniels> from the bisect notes, there are two failure modes

18:34 <warpme_> alyssa: yeah. believe : i was only blindly following git commands....

18:36 <warpme_> tomeu: re: black screen: indeed current master gives "tiling-like" distortions - but it doesn't mean that - at some stages of mesa development - mesa wasn't giving black screen on t720.....

18:37 <tomeu> hrm

18:37 <warpme_> daniels: yes. But black screen i'm interpreting as short phase in time where t720 issue was black screen -not tiling-like distortions....

18:38 <warpme_> in time->in mesa development timeline

18:48 <tomeu> guess some SFBD problem, but wonder why deqp didn't catch it

18:58 <alyssa> warpme_: tiling?

18:59 <alyssa> Interleave modes probably broken

19:00 <warpme_> ehh - i'm describing "tiling-like" just to describe visual distortions seen on screen :-) Thuis is my added value to this project :-p

19:11 <tomeu> wrong tiling stride?

19:16 <daniels> tomeu: surfaceless probably never hits linear render targets

20:25 <alyssa> moo with me

20:25 <alyssa> moooooo

20:25 <alyssa> mooooooooooo

20:25 <alyssa> come on, sir, moo with me!

20:25 <alyssa> 🐮

20:25 <alyssa> 🐄

20:25 <alyssa> moooooooooooooo

20:25 <alyssa> 😂

20:25 <alyssa> thank you.

20:31 <anarsoul> alyssa: what was that?

20:33 <xdarklight> maybe a creative way to say "good moo moooo mooooorning"

20:43 <alyssa> anarsoul: moooooooooooooo

20:43 <anarsoul> xdarklight: I'm pretty sure it's afternoon for alyssa

20:45 <alyssa> 🇨🇦

21:20 <ente> moo?

21:25 <ente> https://jaycarlson.net/embedded-linux/ reading this atm, maybe it's interesting for someone else here

22:34 <alyssa> ente: thank yu.

22:34 <ente> moo!

22:56 <urjaman> moo'n

23:14 <ente> can you drop me off?

23:14 <HdkR> There's a bunch of dorks in here

23:24 <alyssa> 😸