00:05mhenning[d]: snowycoder[d]: As a heads up, the dataflow stuff just landed in https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35446 so you might want to rebase your MR on main whenever you get a chance
00:05mhenning[d]: The syntax for using it is pretty different but the interfaces are the same at the core, so I don't think it will be too painful to update. Let me know if you have any issues with it.
00:08snowycoder[d]: mhenning[d]: Thank you! The new syntax is even cleaner!
01:17gfxstrand[d]: I wonder how many of the WSI issues just come down to "Oops. We didn't use Kopper."?
01:17gfxstrand[d]: I mean, the old path should work, in theory, but still...
02:20mhenning[d]: yeah im secretly hoping we'll turn kopper on and it will magically fix every bug
02:23airlied[d]: in theory we can test that with MESA_DRIVER_LOADER_OVERRIDE=zink
02:25mangodev[d]: mhenning[d]: i hope
02:25mangodev[d]: my current wsi is not great :/
02:25mangodev[d]: if the gpu is under *any* contention, the cursor plane lags, and if under enough stress, even the compositor itself
02:25mangodev[d]: which is why i don't touch most steam games on nvk
02:27mangodev[d]: i hope it's just some jank going on that's an easy fix for the people who work on the driver, and not a deeper rooted issue with zink
02:27mangodev[d]: i have seen improvement though
02:27mangodev[d]: firefox crashes *less* now
02:27mhenning[d]: That may or may not be wsi
02:28mangodev[d]: although in the past week, i've noticed these block artifacts on any page with a `blur` filter in webrender
02:28mangodev[d]: which is strange, as it didn't happen before
02:28mangodev[d]: hard to capture since it only happens for one frame
02:29mhenning[d]: have you filed bug reports? things reported in that channel tend to get lost if they're not an easy fix
02:29mhenning[d]: webrender has a test suite which might help with reproducing that bug
02:29mangodev[d]: mhenning[d]: yeah, but i only file bug reports on things with concrete logs and such
02:29mangodev[d]: i don't want to bloat the issues page with stuff that isn't easily followed
02:30mangodev[d]: mhenning[d]: i hope i could find what causes firefox to crash too
02:30mangodev[d]: because it's definitely webrender related (as it doesn't happen with gpu accel off)
02:30mangodev[d]: although the stack trace dump says it's webextensions for some reason
02:31mangodev[d]: (the dump in journalctl, not the web report)
02:31mangodev[d]: and dmesg has never given me anything of note
02:31mangodev[d]: i check it when filing issues, but it often never has any logs at all related to mesa, only with sata
02:32mangodev[d]: also something else to note is that the discord crash has been fixed for me, i haven't spotted it crashing in a good couple weeks
02:33mhenning[d]: mangodev[d]: I mean, even if it's something that's vague, it's more likely to be fixed as an issue report than an irc message
02:33mangodev[d]: mhenning[d]: i think it's wsi because native wayland windows seem to lag the desktop more than xwayland does
02:34mangodev[d]: i've tested with vkcube, glxgears, and minecraft because they're all easily modifiable
02:34mangodev[d]: although i'd love to dive deeper into test suites, proper benchmarking and the like
02:34mhenning[d]: Is that lag an issue you have with the nouveau gl driver too, or is nvk worse in that regard?
02:35mangodev[d]: mhenning[d]: i've had zink forced the entire time i used nvk in general (also because the commit to force it on was made less than a week after i first started using it :P)
02:36mangodev[d]: i only ran nouveau the first couple boots, and it was slow and unstable
02:37mangodev[d]: it'd be neat to try wlroots's vulkan compositor and see if the issue is there, but it's feature incomplete and i'd probably break my system configuration in trying to set it up
02:37mangodev[d]: (my test pc is my daily driver)
02:38mangodev[d]: on a different note
02:38mangodev[d]: any idea on why building mesa errors with rustup, but doesn't with `extra/rust` (system-packaged rust)?
02:39mangodev[d]: i've been dabbling in rust lately, so i'm still very new to it and trying to learn the ropes (testing the waters, coming from zig and c++)
02:39mhenning[d]: filed a bug for the webrender blur https://gitlab.freedesktop.org/mesa/mesa/-/issues/13459
02:40mhenning[d]: you said that was an issue with i686, right?
02:40mhenning[d]: does your rust compiler in general have i686 support? not sure if that's installed by default
02:40mangodev[d]: mhenning[d]: i think so
02:41mangodev[d]: mhenning[d]: is there a way i can check?
02:41mhenning[d]: I think it used to be you needed to install the stdlib separately for cross compile targets but some of that might have changed
02:42mhenning[d]: mangodev[d]: can you compile a simple "hello world" program for i686?
02:42mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1390523111509852231/image.png?ex=68689110&is=68673f90&hm=38d2e1f2a9470ecc1f635691a8aa479a72623945b757926e6239c8d2640fea23&
02:42mangodev[d]: mhenning[d]: that's a red flag
02:43mangodev[d]: would `rustup target add` help?
02:43mhenning[d]: I don't know tbh
02:43mangodev[d]: wait
02:43mangodev[d]: `rustup target list` lists a ton of built targets
02:44mangodev[d]: OOOooooohhhh i see
02:44mangodev[d]: only one is installed
02:44mangodev[d]: good to know
02:44mangodev[d]: the arch package probably pre-installs the i686 toolchain as well
02:45mangodev[d]: i probably need `i686-unknown-linux-gnu`
02:46mangodev[d]: yay it installed :D
02:46mhenning[d]: I think arch has a different package for it
02:46mhenning[d]: https://archlinux.org/packages/extra/x86_64/lib32-rust-libs/
02:47mhenning[d]: but yeah you might just need the rustup equivalent
02:47mangodev[d]: that may just be the libraries and not the pipeline? unless the libs *are* the toolchain
02:47mhenning[d]: (I don't really know how rustup works)
02:47mangodev[d]: i'm just using it because afaik it's good practice
02:47mhenning[d]: mangodev[d]: The description says "32-bit target and libraries for Rust" so I think it's both
02:48orowith2os[d]: mangodev[d]: Usually is, yeah, at least for developing with Rust. For packagers/distributions/etc, system-packaged rust is fine.
02:48mangodev[d]: i've dealt with java before, so i'm at least *aware* of the benefits of a version manager like `archlinux-java`, i'm just not super sure on to what degree it applies to rust (although probably more to me than a package manager that supports multiple versions of installed packages)
02:49orowith2os[d]: orowith2os[d]: The problems really come when distributions try and ship Rust *dependencies* as packages.
02:50orowith2os[d]: *especially* when it involves forcing them to dynamically link *cough cough* Fedora *cough COUGH*
02:55mangodev[d]: does llvm count?
02:56orowith2os[d]: LLVM doesn't use the rust abi
02:56mangodev[d]: oh
02:56mangodev[d]: you mean the other way around
02:56mangodev[d]: that sounds painful
02:56orowith2os[d]: It is
02:57orowith2os[d]: If ANYTHING in the Rust setup changes, anything at all, the entire set of them has to be recompiled
02:57mangodev[d]: any advice for getting more comfortable with rust? i'm quite new to the language and don't know a lot of etiquette and good practice
02:59mangodev[d]: i've been running through the rust book and rustlings to get a good idea on using the language, but not a ton else other than my previous dabbling with cargo and rustdoc
03:14orowith2os[d]: mangodev[d]: Write a tool or two
03:14orowith2os[d]: Script something
03:14orowith2os[d]: Karol, me, and I'm sure many others here would be able to answer any questions you have
03:14mangodev[d]: fair point
03:14mangodev[d]: i often feel like i learn programming too slow because i don't know what to *use* it for
03:14orowith2os[d]: Rust does VERY nicely for scripting, if I'm speaking my own opinion
03:15orowith2os[d]: I like automating stuff with it
03:15orowith2os[d]: I know there's also a shebang somewhere to just hand out a plain rust file and run that
03:16mangodev[d]: i *adore* low level dev, zig is my dream language after all, but i need to get fully *comfortable* with those parts of programming to be *good* with them
03:16orowith2os[d]: Do you have a microcontroller somewhere?
03:16mangodev[d]: kind of
03:16mangodev[d]: was actually gonna bring up that i have an rpi zero 2w with a touchscreen, and don't know what to do with it
03:17mangodev[d]: it'd be neat to try both zig and rust on it and see where their strong points are, but again, i have no idea what to *do* with it
03:19orowith2os[d]: Simple fading drawing thingy? Diagnostics?
03:19orowith2os[d]: Random pixels?
03:19mangodev[d]: fair point
03:19mangodev[d]: i think i'm often too concerned over practicality
03:20mangodev[d]: i need to pick up the mindset of a 4yo showing their mom their play-doh abomination
03:20mangodev[d]: not concerned about "practicality" or "utility"
03:20mangodev[d]: but rather "i did a thing"
03:20orowith2os[d]: Blink an led
03:21mhenning[d]: yeah, I definitely learn things easier when I have concrete uses for it
03:22mhenning[d]: I initially taught myself how to program by writing games (tiny, not very good games)
03:22mangodev[d]: mhenning[d]: most of my problems are far out of my reach of fixing though
03:22mangodev[d]: most of them are deep-founded or heavily work in progress by folk much, *much* smarter than i am
03:22mhenning[d]: but microcontrollers can also be fun
03:22mangodev[d]: mhenning[d]: same
03:23mangodev[d]: mhenning[d]: i like working with constraints
03:23mangodev[d]: i don't like how powerful modern hardware makes many forget the concept of optimization entirely
03:24mangodev[d]: i likely got my hatred of poorly optimized programs from running a relatively low-end pc for a while
03:24mangodev[d]: started with a dell optiplex with a core 2 quad(?), then moved to an i7 3770k for a while
03:24mangodev[d]: both could only barely run windows 10 itself
03:27mangodev[d]: idk
03:27mangodev[d]: i feel hyperfixated on optimization a lot of the time
03:27mangodev[d]: maybe i'm too obsessive over it, not sure
03:27gfxstrand[d]: mangodev[d]: If it's reproducible enough, a bisect would help greatly. Sometimes if we just know what patch caused it, we can sort of redo the code review on that patch and catch the bug.
03:29mangodev[d]: gfxstrand[d]: i could maybe try, but i don't have a test machine, so i'm worried about screwing over my pc (nvk is my daily driver because proprietary and open kernel module run horridly on the desktop on my setup; ik it's "bad practice," but it's the best i got)
03:29mangodev[d]: maybe some extra specificity could also help?
03:30mangodev[d]: it's always a square artifact in the top-left corner of the screen (no matter *where* the blur is on the page)
03:30gfxstrand[d]: mangodev[d]: This is legitimately really hard. I struggled a lot with that when I was learning. I'd do tutorials and stuff but, like, who actually wants to write a shitty calculator app that's always going to be worse than the one that's already sitting there?
03:31mangodev[d]: feels like an even multiple texture size
03:31mangodev[d]: either 512² or 768²? hard to tell because it's such a brief flicker
03:31mangodev[d]: and i can't play back a video because vlc instantly crashes upon trying to play a video
03:31mangodev[d]: gfxstrand[d]: AGREED
03:31mangodev[d]: also tutorial hell
03:31mangodev[d]: used to fall for that in my gamemaker days, though i eventually got out of it and learned from it
03:32gfxstrand[d]: My first major project was a bit of python that would split up a Blender render and run it on multiple machines and then put all the pieces back together. I was really into Blender at the time and was tired of how long things took. (This was like 2006.)
03:32gfxstrand[d]: Then I started having more fun with my python program than I was having with Blender. 😂
03:35gfxstrand[d]: mangodev[d]: VLC crashing also sounds pretty reproducible. Mind filing a bug with details?
03:36mangodev[d]: gfxstrand[d]: i could actually
03:36mangodev[d]: gave an error about failing to create a wayland surface, nothing to do with video accel
03:38mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1390537140978454638/image.png?ex=68689e21&is=68674ca1&hm=6d549180818421236b050e28389d5dc9423c66c35b926e1ab34f046864720942&
03:38mangodev[d]: it's pretty nondescript ðŸ˜
03:38mangodev[d]: nothing in journalctl
03:39mangodev[d]: nothing in dmesg (as usual)
03:39mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1390537564133392474/image.png?ex=68689e86&is=68674d06&hm=703a3e02ce3f312fa0d706e983ad5b6a92d459582883dd2a6094970ad9c9511b&
03:39mangodev[d]: this is probably concerning, but that i can deal with later
05:15i509vcb[d]: orowith2os[d]: I find that I bash scripts are sad to write where as I can do even the most budget Rust "script" very quickly
05:16i509vcb[d]: mangodev[d]: I find my work with heavy constraints is more masochistic than anything lol
05:17i509vcb[d]: I might be working on an audio codec library for microcontroller targets...
05:18i509vcb[d]: The unique bit is treating all of the heavy signal processing calls as asynchronous since one of my targets has a DSP that I can shove tons of stuff through quickly
05:18mangodev[d]: i509vcb[d]: i mean
05:18mangodev[d]: *heavy* constraints start getting masochistic at a point
05:18mangodev[d]: but working with *tangible* constraints is fun
05:26mangodev[d]: or at least working with some sort of target (e.g. memory usage target, low allocation, cpu time, etc)
07:37luc: hi, i wonder if there's a reason why nova requires drm to be compiled in kernel rather than as a module (depends on DRM=y, instead of just depends on DRM)
07:59jannau: luc: the rust build system requires that at the moment. the rust abstractions are built-in
08:09phasta: luc: this might have to do with some Rust safety requirements (modules can be unloaded). But I'm not sure. From the commit message, it looks as if that might be WIP
08:20OftenTimeConsuming: Hi, I have a 780 Ti msi card and Xorg crashes every time there is a resume from suspend; https://termbin.com/iiat
08:30notthatclippy[d]: Is the DRM=y requirement for nova-core or nova-drm?
08:33jannau: nova-core shouldn't depend on DRM so it should be just nova-drm
08:35jannau: phasta: it's just how the rust abstractions are currently build. the kernel crate is built-in and so all dependencies have to be built-in as well
08:37jannau: I believe the plan is to have drm-rust which can be built as module if DRM=m but that requires rust build changes (already planned)
08:47phasta: Allright, so it's WIP :)
08:55luc: jannau: thanks for your explaination
12:37mangodev[d]: found something i might be able to make an issue about
12:38mangodev[d]: i think there was a zink or nvk regression that heavily affects intellij (testing idea, not sure about other intellij ides)
12:38mangodev[d]: HARD freezes on ANY scrolling events
12:38mangodev[d]: so hard that the entire system locks up for a second or two
12:40mangodev[d]: uhhhhhhhh wait
12:40mangodev[d]: nevermind?
12:40mangodev[d]: updated it and now it's a lot better
12:40mangodev[d]: strange
12:41mangodev[d]: this time i launched through toolbox instead of app launcher though
12:41mangodev[d]: every time i almost find a bug ðŸ«
12:42mangodev[d]: it's like a poorly written episode of scooby doo
12:48zmike[d]: try updating mesa
12:53mangodev[d]: mangodev[d]: .
12:53mangodev[d]: the issue was resolved
12:54mangodev[d]: just as i thought i had something :'
12:55zmike[d]: I fixed it
12:55mangodev[d]: i'm also on last night's build of mesa
12:56mangodev[d]: only new things i see is docs and stuff for other drivers
12:57mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1390677789589377135/image.png?ex=6869211e&is=6867cf9e&hm=c543e79379cb4a84cd3f39cbc65003ef32cefbe384d1020000a6f49a9d8c4bbe&
12:57mangodev[d]: i have these in my current build
13:06zmike[d]: yeah that's what fixed it
13:06zmike[d]: since I assume you're using GL apps
14:11mangodev[d]: something concerning is that today's commits feel slower
14:11mangodev[d]: by a good margin
14:11mangodev[d]: idk how to describe it, but it feels like things slow down easier
14:12mangodev[d]: and discord's doing the flicker thing again
14:12mangodev[d]: meaning it's soft crashing again
14:12mangodev[d]: maybe it just does it less? not sure
14:12mangodev[d]: can't say it's a regression, some boots just feel better than others
15:20gfxstrand[d]: Set `MESA_LOADER_OVERRIDE=zink`.
15:20gfxstrand[d]: Yes, it's already using Zink but you're getting bad window-system paths because things are crazy in the EGL code. That should improve things.
15:21gfxstrand[d]: zmike[d]: Do we have Kopper for GLX? Or is it only EGL?
15:25zmike[d]: it works for everything
15:26zmike[d]: always has
15:26gfxstrand[d]: lkay
15:33asdqueerfromeu[d]: gfxstrand[d]: *`MESA_LOADER_DRIVER_OVERRIDE`
15:34gfxstrand[d]: yes, that
15:35snowycoder[d]: nvk+zink pixmark on my GTX 770 shows artifacts :/
15:35gfxstrand[d]: 😢
15:35gfxstrand[d]: But not on your kepler B?
15:35snowycoder[d]: Nope
15:35gfxstrand[d]: Oh, boy...
15:37snowycoder[d]: I didn't run deqp on it though
15:37snowycoder[d]: but hey, 1.
15:38snowycoder[d]: 1.5x performance on furmark from instruction latencies on GTX 770
15:38gfxstrand[d]: :transtada128x128:
15:39gfxstrand[d]: I wonder if that's just because bigger GPU is better at hiding latency or if we're still getting punished
15:40snowycoder[d]: furmark had lower gains on my GT710 too compared to pixmark, (don't remember how much exactly)
15:40ermine1716[d]: I've booted fedora 42 on RTX 3050, and there are messages like "DP-1 invalid native reply 0x03". Is it known issue?
15:41ermine1716[d]: (That
15:41ermine1716[d]: (That's kernel 6.14.0 though)
15:43gfxstrand[d]: I'm gonna regret refactoring the EGL code, aren't i?
15:43asuasuasu[d]: for whatever it's worth, i have a GTX 760 lying around, which i wouldn't mind donating to whoever in the EU (assuming shipping isn't too much of a pain, and i motivated myself enough for it...)
15:44asuasuasu[d]: ... which should not be broken hopefully
16:24anarsoul: airlied: so far no issues with msleep() workaround on 6.16-rc. It's been running for over a week
16:49mohamexiety[d]: ermine1716[d]: skeggsb9778[d] this sounds like my issue
16:50mohamexiety[d]: for reference ermine: https://discord.com/channels/1033216351990456371/1034184951790305330/1382761134448312404
21:20gfxstrand[d]: Okay, so kopper isn't quite as hosed as I thought it was. We are getting kopper for X11, but in the worst possible way.
21:20gfxstrand[d]: We aren't getting it for Wayland