00:57 <tchebb> Okay, I've confirmed that the gpu_usecount check is not the issue. Added a WARN_ON before the only place it ever gets incremented to ensure the BO isn't on the purgeable list at that time (which is the only reason a race would happen), reproed the crash, and it never fired.

01:10 <tlwoerner> alyssa: ah, i see. thanks. i can add those to my build

01:14 <alyssa> tlwoerner: better if we can get them merged :>

01:14 <tlwoerner> that's always the challenge, n'est pas?

01:16 <alyssa> sigh

01:29 <tlwoerner> at least your pull request is getting some attention :-S

03:29 <icecream95> Looks like there is a memory leak somewhere, as sway is now using 400M of RAM...

03:29 <alyssa> o_o

06:30 <tchebb> icecream95: are client buffers counted towards sway's memory usage? If so, couldn't that be 400MB just from all the surfaces if your display is large enough

06:31 <tchebb> (I am very fuzzy on Linux memory accounting, so that might be a stupid question)

06:35 <HdkR> Easy way to test that would be just to slam open a few dozen alacritty instances

06:46 <tchebb> Okay, very unscientific test: spammed "ps -O rss <swaypid>" and took the lowest reading (~223MB). Opened a new Alacritty instance and did the same, getting ~234MB. So that's an 11MB difference. My screen is 1536x2048, call it 1510x2040 if you ignore Sway's borders, tabs, and bar. 1510*2040*4 == ~12MB framebuffer. So it's at least on the same order of magnitude.

06:46 <tchebb> And the usage went back down after I closed Alacritty.

07:05 <icecream95> tchebb: There's still a difference of a couple hundred megabytes between sway that's been running for a while with two konsole windows, and sway that's just been started with two konsole windows

07:05 <tchebb> ah, fair enough

07:34 <icecream95> macc24: I've got my own kernel working now. I just hadn't included the IT6505 driver when rebasing on 5.10.1

08:13 <macc24> icecream95: where?

08:16 <icecream95> macc24: Where what?

08:16 <macc24> where kernel

08:17 <icecream95> /dev/mmcblk0

08:18 <macc24> ok

08:18 <macc24> where source for kernel?

08:29 <icecream95> macc24: Since you asked, https://gitlab.freedesktop.org/icecream95/duet-kernel

08:33 <macc24> d-does sound work?

08:35 <icecream95> alsamixer crashes with "cannot open mixer: No such file or directory", so I guess not

08:35 <bbrezillon> icecream95: oops, bad conflict resolution => https://gitlab.freedesktop.org/icecream95/duet-kernel/-/blob/master/01-patch-duet.patch#L1704 :)

08:36 <macc24> bbrezillon: that's probably leftover from me merging at middle of night

08:39 <macc24> icecream95: can i steal....... i mean borrow the patches for cadmium?

08:41 <icecream95> macc24: Only if you give them back when you're done with them

08:42 <macc24> yeah, sure just grab them from git repo when you want them

08:43 <macc24> thanks for cleaning it up

08:44 <macc24> there's probably a bunch more changes that could be removed in random spots

08:47 <icecream95> I've got squeekboard compiled now, making tablet mode slightly less useless, but I do miss being able to swipe between the letters like in Chrome OS

08:47 <macc24> try phosh

08:47 <macc24> it runs quite well for me

08:47 <macc24> also manjaro-arm rootfs for lomiri is work-in-progress

08:48 <icecream95> macc24: It seems phosh doesn't work with elgoind yet

08:48 <macc24> icecream95: postmarketos has it working

08:48 <icecream95> Maybe I just haven't installed enough of Gnome to make it work

08:49 <macc24> when installing voidlinux for regular use, should i use platformfs or rootfs?

08:51 <icecream95> macc24: You'd want busctl call --user sm.puri.OSK0 /sm/puri/OSK0 sm.puri.OSK0 SetVisible b true

08:51 <icecream95> oops sorry, copy-pasted wrong clipboard

08:51 <macc24> useful anyway

08:51 <icecream95> void-aarch64-ROOTFS

08:58 <macc24> https://bpa.st/GEVQ

09:01 <icecream95> macc24: What does line 6 do?

09:01 <macc24> icecream95: nothing, i forgot to remove it

09:10 <macc24> icecream95: i think i should tell you one thing, something was up with mali reclocking

09:10 <macc24> and it wasn't bumping the clock up, so i removed few mali's opp-s to force minimal clock higher

09:47 <macc24> icecream95: https://i.imgur.com/RvYH0RH.png

09:53 <icecream95> macc24: Next, you can add Hannah Montana Linux

09:53 <macc24> nah

09:53 <macc24> ill do some emmc improvements

09:54 <macc24> https://github.com/Maccraft123/Cadmium/blob/master/fs/build#L40

09:54 <macc24> this was simpler than i thought lol

12:00 <br_> robmur01: If I simply clear that page fault interrupt (the one that comes without any reason) and return then MMU 'eats' my page tables fine.

12:01 <br_> So I have the first job processed with no errors

12:01 <robmur01> bizarre... :/

12:54 <macc24> icecream95: i have an idea why void didn't work. do you have mkimage and vbutil_kernel?

17:31 <alyssa> https://gitlab.freedesktop.org/alyssa/mesa/-/jobs/6266436

17:31 <alyssa> how do I get the list of fails..?

17:31 <alyssa> hitting browse is error 500

17:34 <daniels> alyssa: download, not browser

17:34 <daniels> *browse

17:34 <daniels> sorry, it is annoying, but it's not the largest fire on my list rn

17:34 <alyssa> ack

17:35 <daniels> (the underlying infra is currently an elaborate three-cup trick which changes every week, so I'm expecting that failure will go away some time over Christmas and be replaced with new and exciting failures)

17:35 <alyssa> weee

17:35 <alyssa> the download just has the log displayed there too

17:36 <alyssa> I mean I guess I could grep our the fails but I remember there being a nice list before the RiiR of deqp-runner

17:36 <daniels> merry festive season one and all

17:36 <alyssa> and I thought the RiiR was supposed to include a nicer list? but maybe lava isn't downloading it

17:36 <daniels> that's one for anholt I guess

20:53 <icecream95> macc24: https://gitlab.freedesktop.org/icecream95/duet-tools

20:55 <macc24> icecream95: are you aware that iio-sensors-proxy is a thing?

20:56 <macc24> icecream95: also i did some cadmium stuff today, and void linux is almost installable on internal emmc like debian

20:57 <macc24> anyway, i've been awake for way to long at this point, ill just go to sleep now.