2022-12-02
00:08chadmedscardracs: like i said yesterday, i WAS working on shipping a bootable usb image but have more pressing matters to attend to right now
00:18chadmedi got as far as booting but it panics upon trying to mount the root. it doesnt like how the vanilla gentoo iso sets up a ramdisk and tries to mount it without a filesystem
00:40beeblebrox has quit [Quit: Konversation terminated!]
01:08asie has quit [Server closed connection]
01:08asie has joined #asahi-alt
01:32marcan< ashi> After yesterdays stream it did work - and mesa was then also no problem to build and run
01:32marcan< ashi> j`ey: well, I was trying to run the driver on M2 Air, and it crashed even though it was supposed to be working
01:33marcanbecause they are known issues and it's you who missed the important bit to make it run, you aren't helping anyone with the report
01:35marcan"It works on the developer's machine with just the right setup because things are still variously broken" is *not* a suitable state for people to go around testing code in
01:35marcanbecause it would involve writing a bunch of documentation to replicate that state, and that in and of itself is wasting our time when the issues are known and will be fixed
01:35marcan has quit [Server closed connection]
01:35marcan has joined #asahi-alt
01:37marcanbut at the same time, you are now using and pointing others towards a random dev branch, *again* (the oq thing)
01:37marcanbecause that's the only one that has the UAPI bump officially, but that doesn't mean it's meant for user testing or intended to be in any way stable
01:38marcanwe're all working on this stuff, we know what we have to do and what the problems are and when things will be ready for user testing
01:38marcantrying to use things before user testing and complaining they're broken is literally just wasting our time with things we already know
01:39marcanbut you are not welcome to report issues in the second case, because we didn't ask for that feedback, because it is not useful
01:39marcanand you are not welcome to encourage others to do so, because you are just making the problem worse then
01:52marcanif this stuff keeps happening, then yes, you're going to force everyone to switch to private repos for development
01:53marcanand so if you're interfering with that and adding nothing of value, you are causing a problem
01:56tpw_rulesdoes this mean you want alt distro folks keep beta package build scripts and so forth private, or just not widely available? for example earlier you had said distros should get ready to build the kernel with rust support. it's natural to want to test the gpu stuff with that
01:57mixi has quit [Server closed connection]
01:57mixi has joined #asahi-alt
01:57marcanthat sounds like something that would be well covered by a single build/run test by the developer, and benefit nothing from wider user testing, so yes
01:58marcanif you are pushing the gpu stuff to users, you aren't getting them to test "rust support"
01:58tpw_rulesit's not clear to me that mps published the APKBUILD in their official capacity as an alpine contributor. that would be an obvious no-no. just letting others know it exists doesn't seem like a huge problem to me
01:59marcantpw_rules: if the APKBUILD is of that branch which as mentioned is a completely random dev branch not intended for *any* user testing whatsoever, then yes, its existence is bad
01:59marcanthat branch was just <alyssa> lina, can you add oq stuff to the uapi? <lina> sure, here you go
02:00marcanand if people can't handle exchanges between developers happening in the open without swooping in and trying to *package* (even not widely) the git branches involved, then yes, maybe we should switch to private exchanges
02:01tpw_rulesso it sounds like you are (justifiably) unhappy with distro folks publically making it easier for third parties to run stuff they can't usefully contribute to at all
02:04marcanmore specifically, stuff bug reports/user feedback for is highly unlikely to be useful at all
02:06alyssa has joined #asahi-alt
02:06marcanbecause people *do* exist that can play around with dev repos like that and still make useful contributions
02:06marcanbut they are rare and need to be intimately familiar with the state of things or make themselves so
02:07marcanand so it's unfortunate if we have to make the choice is between getting a bunch of noise and distractions, and making it impossible for those people to help out at all
02:08marcanand you can help us avoiding having to choose the former by not being part of the problem
02:09alyssafor full disclosure, I have a monitor that marcan purchased for me with project funds (a few hundred CAD$) and that's it
02:11chadmedi was gonna publish what ive done so far on speaker safety today but now im not confident i wont be bombarded with dms from angry people who melted their tweeters :P
02:12chadmedat the same time this is my first real attempt at anything of consequence with rust so some peer review is sorely needed
02:15chadmedi did a stint in the entertainment industry facing both customers and talent, nothing phases me anymore heh
02:16alyssaI have lina on speed dial I get what it's like dealing with talent in the entertainment industry ;)
02:19alyssawith performance rivalling the blobs (despite small team sizes and scarce documentation),
02:19alyssawith feature lists that keep growing (desktop GL? vulkan? opencl? direct3d? yes, I get requests for all of them on a regular basis)
02:20alyssawith demands that I solve their issues when random games that I don't have access to don't work well for them on their underpowered hardware that was never designed to run said content in the first place
02:22alyssabecause being everything that FOSS expects of me is impossible and I'd burn out in days trying
02:23alyssaI don't drop everything when someone files a bug against my software, unless they've put in the work themselves or I have an interest in the problem otherwise
02:24alyssa(If it's for software that I couldn't even run because, e.g. it's a console emulator and I have no legal way to obtain ROMs, then why would I be personally motivated? and if it's not coming from a customer request, why would it preempt the requests that *do* come from paying customers?)
02:25alyssaI don't work in public anymore, I keep code on my local disk until it's ready to see the world, because otherwise people start hitting bugs with that.
02:27alyssaI don't usually respond to private messages, hell I've removed my e-mail from my website because I like my privacy
02:27alyssaI keep wondering what it means for me to be successful in FOSS, because the goal posts keep moving.
02:27alyssaFirst it was "run GNOME on Mali-T760/T860 with no blobs", and that was MY goal. MY definition of success.
02:28alyssaI had a Chromebook and I wanted to run my favourite desktop on it and I needed the Mesa driver to do it. That was it.
02:30alyssaThere was another huge milestone for our team, our driver became officially conformant for OpenGL ES 3.1 on Mali-G52
02:31alyssaA modern API, supporting every OpenGL ES feature that the hardware can do architecturally, stable and correct enough to survive the conformance tests (which are hell, like, 10 hours of GPU driver torture test hell), fast enough to run whatever desktop you want and Neverball and SuperTuxKart without a hitch
02:33alyssaOnce upon a time, if you used Linux on these boards, you had software rendering -- no GNOME for me back in the day -- else you fought to try to find a versioned blob that exactly matched your GPU and OS and glibc version and windowing system and system-on-chip vendor and kernel version and hoped it would work
02:33alyssaNow you install whatever distro you want and you get 3D working out-of-the-box without any blobs and you have no idea you're using our software
02:34alyssaIt's also terrible, because it means the only time people give our software any thought is when it breaks, and then they submit a bug report
02:35alyssaIf I'm lucky, it'll be a recent regression with an accurate bisection, a reasonable way for us to reproduce, and a polite request to fix
02:35alyssaIf I'm unlucky, it'll be a bunch of complaints and entitled demands and little actionable.
02:37alyssaAnd while I do get paid to work on Mali, I don't get paid to triage bug reports that are so far out of scope of anything this hardware was intended for that it's not clear a priori whether there's anything that can or should be done at all.
02:38alyssaFor AGX I'm setting the boundary that I will not be looking at any bug reports other than regressions, and I expect due diligence on those (ideally bisection)
02:38alyssaPeople can still open bugs, I literally cannot stop you (it's a public Mesa tracker and not one I can opt out of)
02:39alyssabut a public bug tracker is not an entitlement to developer time, it's a place to keep records of what wasn't working at one point in time
02:39alyssamaybe somebody else will feel motivated to fix your issue (maybe Lina will if it'll make for a good stream)
02:40alyssamaybe you'll hire a freelancer to fix it (I think Christian Gmeiner was doing this? or maybe Jonathan Marek? one of the etnaviv people had a "I'll quote you for any FOSS bug" setup)
02:40alyssaand maybe it'll never be solved and maybe it wasn't that critical in the big picture anyway
02:41alyssaIf I honour every demand that's placed on my time, if I give credence to every hateful comment people make about me online as recently as this week, then I'll never be successful,
02:42alyssabecause frankly people expect me to a superhuman perfect graphics driver developer with unbounded resources and who never makes a single error, and also be cis male and allistic while doing so
02:43alyssacertainly I don't know a single person who would meet it, and I know just about everyone in Mesa
02:44alyssa(and shit on open source drivers all you want, relative to team sizes Mesa developers are going to be far more productive than our closed source counterparts)
02:45alyssaI want to say "it's also not a standard placed on my male colleagues" although thinking back to Phoronix comments I don't think that's true, there's a lot of entitlement to go around.
02:45alyssa(I can think of a handful of male Mesa developers who the FOSS community puts through similar amounts of crap... It shouldn't happen to me and it shouldn't happen to them.)
02:46alyssaSo circling back... what does success look like for me in FOSS? I think I have to define it for myself.
02:46chadmedmy friends and i have a term for this, "slowmaxxing". set your own goals, measure your success against nothing else, anyone who disagrees can impolitely go f themselves. that said i cant even begin to fathom the level of personal hate you must receive, though im sure it gets difficult to ignore
02:47alyssaI have some future goals for FOSS, the most near ones in the next week and the more ambitious ones years on the horizon
02:48alyssa[ ] Subgoal, learn about SSA-based register allocators. Still working on this one slowly.
02:50alyssa[x] Make sure M1 Linux gets done right w.r.t the graphics stack. I had some horror image of people running Zink on MoltenVK on Apple's Metal blob on darling on Corellium Linux and yeah no let's not do that.
02:51alyssa[x] Run Quake on M1, this was actually a joking bet I made back in early 2021, whether I could run Quake before marcan could write a kernel driver. Technically I won that one thanks to Lina swooping in, marcan still hasn't written a kernel driver for this gpu ;-p
02:51alyssa[ ] Ship open drivers on Asahi Linux on M1. Not just in edge, in the real release. IDK how long it'll be until that happens but it's exciting to think about.
02:55alyssa[ ] Learn how to write better GPU drivers, and how to better write GPU drivers. I'm learning every day, this one. Lot of counterintuitive stuff. Doing a lot better now than I was 5 years ago when I started out, that's for sure.
02:56alyssaBy that metric I'd say I've been successful and if I disappeared from FOSS tomorrow, nobody could take that success from me.
02:58ma4 has joined #asahi-alt
03:00alyssa(this one has been brewing for a while, it's not super related to your rant but I've been meaning to vent this somewhere and my blog/masto are both too high traffic)
03:05ma3 has quit [Ping timeout: 480 seconds]
03:10alyssaso in the future instead of people on those sites wasting their time writing the same comments, they can just say '5' '3' '4' or whatever
04:30qyliss has quit [Server closed connection]
04:30qyliss has joined #asahi-alt
04:41espo has quit [Server closed connection]
04:41espo has joined #asahi-alt
04:54jamespmorgan has joined #asahi-alt
04:57veloek has quit [Server closed connection]
04:57veloek has joined #asahi-alt
04:59jamespmo_ has quit [Ping timeout: 480 seconds]
05:01mjolnir[m] has quit [Server closed connection]
05:01mjolnir[m] has joined #asahi-alt
05:01matrix_ds has quit [Server closed connection]
05:02ahdeb[m] has quit [Server closed connection]
05:02ahdeb[m] has joined #asahi-alt
05:04matrix_ds has joined #asahi-alt
06:32ashi has joined #asahi-alt
06:40ashiI did some comments on asahi-dev, that was not on purpose, I was tired and joined the wrong channel, sorry again. Anyway I won't hang out here anymore for a while, because there is no place for "people like me" here. I am not an end user, I am using linux since the early 90s. Did kernel driver develoment in the 2000s. Know how to fix my stuff, know my debian. I no not need help with grub or apt. The only thing that is interesting
06:40ashito me is playing with dev branches and be happy about improvements knowning that you owe me nothing and things can break. Since "people like me" are not allowed to talk about that here, there is no point in being here actually.
06:46ashi has quit [Quit: Leaving]
07:31mpsmarcan: tpw_rules is right, I created APKBUILD for my local testing and didn't pushed it to official alpine repo
07:33mpsand I just told ashi that I have ready and don't need guide how to build mesa, to not waste time
07:33chadmed_ has joined #asahi-alt
07:36jamespmo_ has joined #asahi-alt
07:42jamespmorgan has quit [Ping timeout: 480 seconds]
08:43kujeger has joined #asahi-alt
09:06sven has joined #asahi-alt
10:05cylm_ has joined #asahi-alt
10:06cylm has quit [Ping timeout: 480 seconds]
10:41chadmed_ has quit [Remote host closed the connection]
10:45chadmed_ has joined #asahi-alt
11:03jamespmorgan has joined #asahi-alt
11:09jamespmo_ has quit [Ping timeout: 480 seconds]
12:06jannaumps: how often in that time did you "test" in random developer branches. I'd argue that random developer branches are not published code
12:09svenfwiw, I don't mind if other *developers* run my broken branches. when jannau e.g. tested the dp altmode thing he figured out why 4 lane DP mode sometimes didn't work because he understands DCP and can debug atcphy.
12:10sventhat requires context thought to differentiate between "obvious missing feature / bug that sven will eventually get to" vs. "bug he just didn't trigger on his hardware"
12:14jannauthat was a reply to "through years on different software projects no one told 'do not test published code'"
12:15mpssure. and I don't like to push to users even released things before I or someone else test them
12:17mpsbut for decades whenever I ask about test things not yet released usually devs says 'please test and report possible problems/bugs/...'
12:17ashi has joined #asahi-alt
12:18svenwhat's beyond that is probably best described as "unfinished and possibly broken hacks shared between developers"
12:18chadmed_the distinction here is that the "dev branches" people have been randomly picking and testing arent even the ones that are suitable for testing, theyre like pre-alpha hacks
12:20chadmed_yeah so taking the pre-alpha stuff that everyone knows is broken and going "hey this is broken" is a waste of time
12:20ashiExactly. And I *am* a *developer* - for decades. Just not an *Asahi developer*. I did not publish or distribute any of your dev branches. Just wanted to share with like-minded tinkeres - which is forbidden, right?
12:21ashiAnd your hacked dev branches, only allowed to be built by other core inner circle asahi devs. Yes, they should be private
12:22_jannau_you are free to test things and fix bugs but we are not interested in bug reports on random developer branches
12:22chadmed_ashi: again, no one is saying you cant _try_ things but just keep in mind that if theres something broken, theres like a 99.9999% chance the developers already know about it
12:23chadmed_and the problem with sharing them publicly in such well-populated chats is that people are eventually going to stumble across them and go "hey cool ${FEAUTRE} that ive been waiting for"
12:23_jannau_also fixing bugs in random branches is probably a bad idea since that easily could be just duplicate work
12:23chadmed_then try to build it and get all huffy and pissed when it doesnt work 100% immediately with no pain
12:23j`eyashi: they've been honest about not wanting people to test (and then report issues) on random branches, seems like a fair and honest thing
12:24ashichadmed_: yes I know it, I did not open an issue on github or spammed the dev channel, I thought asahi-alt would be okay to discuss things to improve the debian experience in whatever way. The instructions and packages provided by Glanzmann are debian TESTING and an -rc kerne, both beta, alpha whatever, so adding more wip to that for testing, did not mean any harm
12:24chadmed_like this literally happened two days ago and it was a horrible frustrating experience for everyone involved. the people who tried to help, the person who was trying to build the branch, everyone
12:25ashij`ey: Yup, but publishing (ie. pushing) code to branches that are not intended to be built.... hmm...
12:26j`eyashi: you are free to build it, but the point is to not report issues etc, if it doesnt work
12:26ashij`ey: okay, so when I said in #asahi-alt "I built it and it does not work" is that a bug report to devs?
12:27j`eyI kinda disagree with sven, I think it depends on the way you say it, if it's like "oh i tried it, ddidnt work for me" vs "help, it didnt work for me, whats wrong" etc
12:29chadmed_but again, recall what happened a couple of days ago with that person who was "asking" for help
12:29ashiAnd I just left it like that until it started to work, I was happy. and expecting ZERO SUPPORT and I know your perspective also being a developer myself.
12:29chadmed_and strongly discouraging the building of random non-canonical branches that are mostly feature incomplete is a good way to avoid it
12:33ashiI am really grateful for everything that works and not a bit annoyed if something doesnt. I was naming the branch to @mps I built, and that was enough for a marcan rant. Don't get it.
12:40ashiOkay so my takeaway is - "do whatever you want with the random branches but don't tell anyone". Does that fit?
12:41chadmed_like ive told some mates about/shown them the state of the art in private and irl but i know theyre not going to jump in here and see a funny link that says "gpu msaa should be working" and click it
12:44beeblebrox has joined #asahi-alt
12:49chadmed_ has quit [Remote host closed the connection]
13:09ashi has quit [Quit: Leaving]
13:11r0ni has quit [Ping timeout: 480 seconds]
14:30alyssaliterally agx/next right now is asahi/main plus a bunch of commits labelled as HACK, WIP, and/or known broken :-p
15:01janrinze has quit [Quit: Leaving.]
15:01janrinze has joined #asahi-alt
20:53benzmac16v has joined #asahi-alt
21:01benzmac16v has quit [Ping timeout: 480 seconds]
21:04jamespmo_ has joined #asahi-alt
21:09jamespmorgan has quit [Ping timeout: 480 seconds]
21:30alyssa has quit [Quit: University awaits]
22:17jamespmorgan has joined #asahi-alt
22:23jamespmo_ has quit [Ping timeout: 480 seconds]
22:50jacksonchen666 has joined #asahi-alt
23:42jamespmo_ has joined #asahi-alt
23:49jamespmorgan has quit [Ping timeout: 480 seconds]