05:53 <endrift> alyssa: I've started getting to the point where if I want to cherry-pick things between branches that I just committed I'll just `git checkout whatever && git cherry-pick HEAD@{1}` because I have become too powerful

05:53 <endrift> this is what reflog let me become

08:00 <HdkR> Got myself an x86-64 panfrost build, now I just need to setup a board to run it

09:09 <tomeu> HdkR: what could go wrong? :p

10:01 <icecream95> alyssa: AlphaEnabled is different from alpha.enabled, so it should have been obvious that there was code in the middle that could be changed...

10:01 <icecream95> Keeping the assert will remind me to add lowering in Nine, for which support should be coming Soon™

10:03 <icecream95> It's not easy debugging x86 code in a buggy emulator when you haven't even got a compiler set up

12:05 <warpme_> guys: I just gave try for current mesa master on mali450/t720/t820/t860/g31 and all works ok except t720. errant t720 (allwinner h6) gives screen like this: https://paste.pics/aa5877a2bea3b996fdaa622ab33d0c88 VNC on this machine however gives screen ok so it seems like issue is close to drm(?) As mesa from few weeks earlier works ok on this machine - i suspect this might be regression?

12:10 <robmur01> hmm, that looks a lot like linear scanout of a tiled format. Possibly related to the recent modifier stuff?

12:21 <daniels> warpme_: which client is that running?

12:21 <warpme_> mythtv redering UI via GLES3

12:21 <warpme_> and zorg 1.20.8

12:22 <warpme_> zorg->xorg

12:22 <daniels> ah, X ...

12:23 <macc24> is segmentation fault when playing newest supertuxkart expected?

12:27 <macc24> mali t760mp4, mesa 20.1.6

12:43 <daniels> macc24: nope

12:43 <daniels> er, actually, it might be the assert fixed in master? I didn't think that had crept into stable but maybe it has

12:44 <daniels> do you have a backtrace?

12:44 <macc24> how to get backtrace?

12:44 <macc24> i just saw "Segmentation fault"

12:45 <daniels> try 'ulimit -c unlimited' to get a core dump, then gdb $(which supertuxkart) core, then type 'bt' at the prompt

12:47 <daniels> (or if you're just running under Wayland or something rather than directly on KMS, you can just run 'gdb supertuxkart')

12:48 <macc24> https://bpa.st/4EIQ

12:53 <robmur01> daniels: you mean not everyone runs native KMS stuff via SSH from another machine? :O

12:59 <tomeu> macc24: doesn't seem to be related to panfrost

13:00 <tomeu> macc24: one would need to see what CallSystemFunctionNative does

13:00 <tomeu> it's maybe trying to call a NULL func pointer?

13:01 <daniels> robmur01: heh

13:08 <HdkR> Null function pointer would show up in backtrace

13:16 <HdkR> tomeu: Expectation is that getting an ARM board setup with the latest kernel will be the hard bit for me :P

13:33 <alyssa> endrift: wonderful

13:33 <alyssa> icecream95: sure :)

13:37 <alyssa> warpme_: right, that definitely looks like modifiers got botched

13:37 <alyssa> I'm unsure why that would affect T720 but not T860

13:38 <alyssa> warpme_: https://people.collabora.com/~alyssa/hack.diff

13:38 <alyssa> ^ If that makes the issue go away, we can confirm it's modifiers, but that doesn't tell us why.

13:40 <warpme_> alyssa: perfect. pls give me some minutes as i'm in middle rebuilding with 5.8.5 kernel (report was from 5.9.rc3). builder should finish in 20-30min....

13:40 <alyssa> sure

13:59 <HdkR> Hmmm, I have a rockpro64 somewhere around here I Think

14:01 <HdkR> alyssa: Is the rockpro64 I have here one you have access to?

14:01 <alyssa> HdkR: Yeah

14:01 <alyssa> currently with kbase on it, I'd appreciate if that one stayed as currently configured

14:01 <alyssa> something something I tried to send you a rp64 already but nope ;P

14:02 <HdkR> :)

14:04 <HdkR> Time to order another one

14:04 <alyssa> what for?

14:04 <HdkR> I want to do Panfrost x86-64 testing and it seems like the best choice

14:05 <HdkR> Bifrost isn't yet at GL 3.x as far as I'm aware :P

14:05 <alyssa> HdkR: Sadly no :p

14:06 <HdkR> Ordered another one plus a couple of cases so I can glue the two boards together to live forever together

14:06 <HdkR> I need to clean up my SBC shelf. It's a huge mess

14:08 <warpme_> alyssa: with this patch app seems to be hanging. screen is black. i see xorg server started (no errors) - but app. is hanging

14:09 <robmur01> if only Atom C3-3445 devices actually existed... plenty of ASUS ZenPads for potential x86-Lima shenanigans, but no lovely T720s for x86-Panfrost :(

14:09 <alyssa> warpme_: ....uh

14:10 <HdkR> robmur01: It's more sane to keep the kernel space ARM and let the userspace play in the x86 pool anyway ;)

14:27 <alyssa> bbrezillon: I'm noticing some serious flushes on manhattan, wondering if we need to batch harder

14:56 <bbrezillon> alyssa: do you know what causes those flushes?

14:56 <bbrezillon> clears maybe?

14:56 <alyssa> bbrezillon: finding out now

14:57 <alyssa> looks like bad AFBC logic isn't helping

14:59 <alyssa> oh, interesting, entirely caused by the AFBC flushes

14:59 <alyssa> (no flushes if I disable AFBC)

15:00 <alyssa> so that's entirely my fault, ignore me

15:09 <alyssa> okay, right -- mh doesn't actually read from AFBC textures, but it still hits the stupid flush to find out if it needs to

15:14 <alyssa> AFBC does bring up some 'interesting' questions ;-;

15:15 <HdkR> :)

15:15 <HdkR> Compression is always fun

15:18 <HdkR> Just make every hardware block compression aware. Easy right? </s>

15:18 <alyssa> =)

15:19 <alyssa> uhhh