07:32 <AreaScout> sigmaris thanks for confirming and the commit link ! :)

08:38 <icecream95> Has the cause of the memory leak in 5.9 been found yet?

08:45 <macc24> the WHAT?

08:53 <icecream95> Apparently it's fixed in 5.9.5, so I'll try compiling that

09:21 <warpme_> Alyssa: pls ping me when You will have sec. I want to recall issue of distorted screen on panfrost on sun4i. it is related to 2a121751 and our chat few days ago....

09:29 <macc24> warpme_: doesn't sun4i have Mali400?

09:36 <warpme_> macc24: this depends on Soc. allwinner h6 has t720 and this is hw i have issue

09:39 <macc24> yeah but sun4i refers to specific SoC. allwinner a10 is sun4i, https://linux-sunxi.org/A10

09:43 <warpme_> ah ok - i was imprecise then. I was referring to drm driver. it is sun4i. issue is about cooperation of DE3 in h6, current panforst and dma_buf modifiers

09:44 <macc24> oh

09:45 <macc24> that explains the confusion

09:45 <warpme_> :-)

11:25 <daniels> warpme_: what about modifiers ... ?

13:12 <warpme_> daniels: sorry for delay (prof.life steals me). Basically: 2a121751 exposes interesting thing (IMHO) with modifies. It is manifesting by distorted screen on allwinner h6 SoC. Reverting 2a121751 fixes. things. Together with jernej we done some tests. Some side notes: DisplyEngine3 used by h6 supports only linear format (from list of mesa known formats). So we see following: 1\ Currently kernel on DE3 H6 hw. reports no

13:12 <warpme_> "IN_FORMATS" property (no list of modifiers). Screen is distorted. 2\add property list with linear modifier as the only member (http://ix.io/2D1V) fixes issue. In my opinion this is bug as modifiers concept it "default+differentiator" type of concept where: if both sides (consumer-provider) are using default - differentiator can be absent OR empty. Current observations - for me - are showing wrong behaviour for

13:12 <warpme_> absence of modifier case.

13:17 <daniels> warpme_: interesting - as you say, adding explicit linear advertisement is correct and helpful, but in the absence of any explicit modifier list, then linear should be returned anyway. which userspace are you seeing this on?

13:17 <warpme_> current mesa

13:18 <warpme_> oh - exactly mesa few days old....

13:20 <daniels> that's ok, nothing's changed in the last few days (ttbomk) which would impact that

13:21 <daniels> when I say 'which userspace', I mean ... are you using Kodi on GBM? Weston? GNOME Shell? Xorg? other?

13:21 <warpme_> i'll give check for todays mesa - just to be sure i'm not false positive

13:29 <warpme_> daniels: oh sorry. it is mesa on xorg for mythtv

13:30 <daniels> ah, Xorg ...

13:30 <warpme_> yeah. i'm really old-fashion guy :-)

13:30 <daniels> is Xorg distorted when it starts up? is it distorted when you run a window manager and if so, which window manager? is MythTV distorted in fullscreen mode?

13:32 <warpme_> distortion is as soon as xorg starts rendering (as it uses GL via glamour - any Xorg output is distorted)

13:34 <daniels> aha

13:40 <daniels> warpme_: does this fix it? https://gitlab.freedesktop.org/daniels/mesa/-/commit/9fc9c15e990215fcaca634c59fd1d3dcac1d59fd

13:40 <warpme_> give me sec....

14:05 <warpme_> daniels: nope. current mesa gives distorted screen

14:10 <warpme_> i mean current mesa + your fix of course

14:30 <daniels> damn :(

14:57 <jernej> daniels: regarding panfrost_query_dmabuf_modifiers callback - shouldn't it check if modifiers is NULL and return only the number of supported modifiers in that case?

14:57 <jernej> at least lima and vc4 do that

15:00 <daniels> jernej: oh yeah, it absolutely should - I left a review comment in one of the MRs about that but it never got caught :(

15:01 <jernej> but I kinda doubt this is causing the issue

15:03 <narmstrong> it shouldn

15:04 <narmstrong> it shouldn't assume a specific modifier from the display controller for sure

15:04 <narmstrong> maybe one day we will have a display controller only supported compressed buffers with modifier

15:05 <jernej> narmstrong: IN_FORMATS is optional property, mesa should assume linear if it is missing

15:17 <archetech> worse case it can be made into a hackintosh next yr

15:17 <archetech> ;p

15:17 <archetech> oops wrg chan

15:40 <italove> chewitt: hi, I'm trying to setup the blob in my RK3399 to debug something and I'm wondering what's the easiest way to go about doing that, someone suggested that you might be able to help with that

16:16 <macc24> italove: why

16:24 <daniels> macc24: to trace the blob and compare it against Panfrost to improve Panfrost

16:24 <macc24> understandable, have a nice day

16:25 <daniels> enjoy your weekend

16:27 <italove> haha

16:30 <daniels> italove: I think it might well be into the weekend for him, as IIRC he's in eastern Europe?

16:30 <macc24> poland is not in eastern europe

16:31 <italove> daniels: idk, but for now I'm trying to setup it like boris said, hopefully it works without much trouble

16:32 <daniels> *central Europe

16:32 <daniels> italove: fingers crossed ...

16:35 <macc24> wait how do you know

17:20 <bbrezillon> italove: FWIW, I just realized my bifrost setup was not working either (mali_kbase+blob-r16p0) :-/

17:21 <italove> bbrezillon: was it working before?

17:21 <bbrezillon> italove: not really

17:21 <bbrezillon> but it was good enough to dump deqp traces

17:21 <bbrezillon> except this time I need something after the first frame :)

17:22 <italove> hmm

17:23 <bbrezillon> tomeu: did you have https://gitlab.freedesktop.org/tomeu/mali_kbase/-/tree/BX301A01B-SW-99002-r16p0-01rel0 to work on a VIM3?

17:26 <bbrezillon> here's my version https://gitlab.freedesktop.org/bbrezillon/mali_kbase/-/commits/BX301A01B-SW-99002-r16p0-01rel0/ (just one change to have it compile on 5.9+)

18:55 <archetech> did ya know this is possible ? I built my xorg with glamor

18:55 <archetech> 35.485] (II) GLX: Initialized DRI2 GL provider for screen 0

18:56 <archetech> accelerated X