ChanServ changed the topic of #asahi-alt to: Asahi Linux: porting Linux to Apple Silicon macs | User-contributed/unofficial distribution ports | Logs: https://alx.sh/l/asahi-alt
<chadmed>
scardracs: like i said yesterday, i WAS working on shipping a bootable usb image but have more pressing matters to attend to right now
<scardracs>
chadmed: sorry, my bad. it has been a long day and I forgot about it
<chadmed>
all good :)
<chadmed>
i 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
beeblebrox has quit [Quit: Konversation terminated!]
asie has quit [Server closed connection]
asie has joined #asahi-alt
<marcan>
< ashi> After yesterdays stream it did work - and mesa was then also no problem to build and run
<marcan>
< 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
<marcan>
I bet you forgot that bti was broken and you had to use arm64.nobti
<marcan>
and *this* is why testing random branches and reporting issues wastes everyone's time
<marcan>
because 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
<marcan>
"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
<marcan>
because 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
marcan has quit [Server closed connection]
marcan has joined #asahi-alt
<marcan>
but at the same time, you are now using and pointing others towards a random dev branch, *again* (the oq thing)
<marcan>
which already caused a bad bug report
<marcan>
because 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
<marcan>
see where the problem is?
<marcan>
we'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
<marcan>
trying to use things before user testing and complaining they're broken is literally just wasting our time with things we already know
<marcan>
it's not "not welcoming", it's just wasting developer time
<marcan>
you are welcome to test things, *after* they've been released for user testing
<marcan>
and you are welcome to do whatever you want on your own
<marcan>
but you are not welcome to report issues in the second case, because we didn't ask for that feedback, because it is not useful
<marcan>
and you are not welcome to encourage others to do so, because you are just making the problem worse then
<marcan>
< mps> ashi: I alredy created APKBUILD for alpine
<marcan>
oh ffs running random branches is bad enough, PLEASE DO NOT PACKAGE THEM
<marcan>
like seriously, you do realize you are just hurting us developers right?
<marcan>
what you're doing is like Manjaro, but worse
<marcan>
if this stuff keeps happening, then yes, you're going to force everyone to switch to private repos for development
<marcan>
if you don't want that...
<marcan>
STOP
<marcan>
we're trying to get full Linux support for a proprietary platform with a tiny team
<marcan>
our goal is to get there as fast as possible with a high quality result
<marcan>
and so if you're interfering with that and adding nothing of value, you are causing a problem
<marcan>
we don't have the time for problems
<marcan>
we want to get this stuff done
<tpw_rules>
does 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
mixi has quit [Server closed connection]
mixi has joined #asahi-alt
<marcan>
that 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
<marcan>
if you are pushing the gpu stuff to users, you aren't getting them to test "rust support"
<marcan>
you are getting them to test the gpu stuff
<marcan>
we haven't even merged the gpu stuff into *our* tree yet
<tpw_rules>
it'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
<marcan>
why would alt distro folks need to do that?
<marcan>
tpw_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
<marcan>
that branch was just <alyssa> lina, can you add oq stuff to the uapi? <lina> sure, here you go
<marcan>
it was an exchange between two people
<tpw_rules>
ok, thanks for clarification
<marcan>
and 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
<marcan>
you want openness, don't turn it into a problem for us
<marcan>
otherwise you won't get openness
<tpw_rules>
so 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
<marcan>
correct
<tpw_rules>
makes sense.
<marcan>
more specifically, stuff bug reports/user feedback for is highly unlikely to be useful at all
<marcan>
because terrible signal to noise ratio bug report channels are not useful
<tpw_rules>
yes
<marcan>
and don't get me wrong, it will be sad if we have to make this stuff private
alyssa has joined #asahi-alt
<alyssa>
I've already switched to all private branches, this was a good move \shrug/
<marcan>
because people *do* exist that can play around with dev repos like that and still make useful contributions
<marcan>
but they are rare and need to be intimately familiar with the state of things or make themselves so
<alyssa>
then again my threat model is different than marcan's
<alyssa>
(well, a superset of marcan's I guess)
<marcan>
and 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
<marcan>
and you can help us avoiding having to choose the former by not being part of the problem
<alyssa>
i don't get paid for asahi i'll enjoy my private branches thanks
<alyssa>
for full disclosure, I have a monitor that marcan purchased for me with project funds (a few hundred CAD$) and that's it
<alyssa>
and that was for DCP anyway
<chadmed>
i 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
<alyssa>
yep..
<chadmed>
at the same time this is my first real attempt at anything of consequence with rust so some peer review is sorely needed
<alyssa>
yep..
<alyssa>
now you know one of the many sources of FOSS burnout (-:
<alyssa>
so.. learning experience! (-:
<chadmed>
i did a stint in the entertainment industry facing both customers and talent, nothing phases me anymore heh
<alyssa>
yeah i get that i mean
<alyssa>
I have lina on speed dial I get what it's like dealing with talent in the entertainment industry ;)
<alyssa>
(more importantly, lina has *me* on speed dial :p)
<chadmed>
ah the emergency red rotary dial phone
<alyssa>
IDK
<alyssa>
people expect me to work miracles
<alyssa>
delivering drivers for hardware I don't even have within months of them being released,
<alyssa>
supporting features that the hardware itself lacks,
<chadmed>
you havent given us a reason to expect otherwise yet!
<alyssa>
with performance rivalling the blobs (despite small team sizes and scarce documentation),
<alyssa>
with feature lists that keep growing (desktop GL? vulkan? opencl? direct3d? yes, I get requests for all of them on a regular basis)
<alyssa>
with 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
<alyssa>
shade thrown at me both in public and in private
<alyssa>
throw in severe code-of-conduct violations and...
<alyssa>
uhh
<alyssa>
yeah, I'll keep my private branches and private bug trackers and private life
<alyssa>
because being everything that FOSS expects of me is impossible and I'd burn out in days trying
<alyssa>
People wonder how I do so much, but the truth is that what counts is what I don't do
<alyssa>
I 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
<alyssa>
(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?)
<alyssa>
I 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.
<tpw_rules>
i think that's entirely fair
<alyssa>
I don't usually respond to private messages, hell I've removed my e-mail from my website because I like my privacy
<alyssa>
I keep wondering what it means for me to be successful in FOSS, because the goal posts keep moving.
<alyssa>
First it was "run GNOME on Mali-T760/T860 with no blobs", and that was MY goal. MY definition of success.
<alyssa>
I 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.
<alyssa>
And... I did it? June 2019?
<alyssa>
And I guess I should have been satisfied there
<alyssa>
but that was GLES2 on v5 Mali
<alyssa>
people wanted GLES3.2, OpenCL 3.0, Vulkan 1.3, Direct3D
<alyssa>
on Mali v4, v5, v6, v7, v9, v10
<alyssa>
on AGX G13, G14
<alyssa>
There was another huge milestone for our team, our driver became officially conformant for OpenGL ES 3.1 on Mali-G52
<alyssa>
like, capital-C Conformance
<alyssa>
and we did it again less than a year later for Mali-G57
<alyssa>
(v7 and v9 respectively)
<alyssa>
Surely, now surely that was a success
<alyssa>
A 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
<alyssa>
what's more, we became *infrastructure*
<alyssa>
Our driver's shipping in virtually every Linux distro, in the default images
<alyssa>
Once 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
<alyssa>
Now 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
<alyssa>
That's awesome!
<alyssa>
It'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
<alyssa>
If 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
<alyssa>
If I'm unlucky, it'll be a bunch of complaints and entitled demands and little actionable.
<alyssa>
I feel bad letting the bug reports languish, but what am I going to do?
<alyssa>
I don't get paid to work on AGX, this is purely hobby.
<alyssa>
And 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.
<alyssa>
For 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)
<alyssa>
People can still open bugs, I literally cannot stop you (it's a public Mesa tracker and not one I can opt out of)
<alyssa>
but 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
<alyssa>
maybe somebody else will feel motivated to fix your issue (maybe Lina will if it'll make for a good stream)
<alyssa>
maybe you'll learn to fix your own issue
<alyssa>
maybe 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)
<alyssa>
and maybe it'll never be solved and maybe it wasn't that critical in the big picture anyway
<alyssa>
There's no free lunch
<alyssa>
So... what is success for me in FOSS?
<alyssa>
If 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,
<alyssa>
because 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
<alyssa>
and that's a standard that's impossible to meet by design
<alyssa>
certainly I don't know a single person who would meet it, and I know just about everyone in Mesa
<alyssa>
(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)
<alyssa>
I 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.
<alyssa>
(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.)
<alyssa>
So circling back... what does success look like for me in FOSS? I think I have to define it for myself.
<chadmed>
my 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
<alyssa>
yeah, that
<alyssa>
I have some future goals for FOSS, the most near ones in the next week and the more ambitious ones years on the horizon
<alyssa>
I won't share those here, mostly because people will hold me to them if I do
<alyssa>
But I will share some past ones
<alyssa>
[x] Run GNOME on RK3288 and RK3399 bloblessly
<alyssa>
[x] Run SuperTuxKart well on RK3399
<alyssa>
[x] Get Mali-G52 to parity with RK3399
<alyssa>
[x] Get Mali-G57 to parity with RK3399
<alyssa>
[x] Learn as much as I can about compiler theory, put what I learn into practice
<alyssa>
[ ] Subgoal, learn about SSA-based register allocators. Still working on this one slowly.
<alyssa>
[x] Draw a triangle on M1 as soon as possible
<alyssa>
[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.
<alyssa>
[x] Run GNOME on M1 flawlessly
<alyssa>
[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
<alyssa>
[ ] 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.
<alyssa>
[ ] " on the M2 Air that torvalds uses...... senpai notice meeeee
<alyssa>
[x] Do the Twitter thing
<alyssa>
[x] Delete my Twitter
<alyssa>
[ ] 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.
* alyssa
shrugs
<alyssa>
Most of those little boxes are ticked off
<alyssa>
By that metric I'd say I've been successful and if I disappeared from FOSS tomorrow, nobody could take that success from me.
<alyssa>
Not entitled Linux users, not online bigots, not abusers, nobody.
<alyssa>
So maybe that's success, then.
<alyssa>
And maybe I'll keep my private branches :)
ma4 has joined #asahi-alt
<alyssa>
maybe the fact so much shit gets thrown at me is a testament to that success, IDK
<alyssa>
marcan: there, you ranted so I did a rant too :p
<alyssa>
(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)
<alyssa>
I wrote this a year ago but it is still so true
<alyssa>
only change is adding a mean comment making fun of Lina :-(
<tpw_rules>
i don't get it. i'm not sure saving those kinds of things is useful
<alyssa>
tpw_rules: parody of the comments about anything I publish on any of { reddit, HN, etc }
<alyssa>
same tropes every time
<alyssa>
hey there's a thought, maybe those comments should be numbered
<tpw_rules>
oh i see
<tpw_rules>
maybe we need that old herp derp for youtube extension for the whole web
<alyssa>
so 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
<alyssa>
would be much more efficient
qyliss has quit [Server closed connection]
qyliss has joined #asahi-alt
espo has quit [Server closed connection]
espo has joined #asahi-alt
jamespmorgan has joined #asahi-alt
veloek has quit [Server closed connection]
veloek has joined #asahi-alt
jamespmo_ has quit [Ping timeout: 480 seconds]
mjolnir[m] has quit [Server closed connection]
mjolnir[m] has joined #asahi-alt
matrix_ds has quit [Server closed connection]
ahdeb[m] has quit [Server closed connection]
ahdeb[m] has joined #asahi-alt
matrix_ds has joined #asahi-alt
ashi has joined #asahi-alt
<ashi>
I 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
<ashi>
to 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.
ashi has quit [Quit: Leaving]
<mps>
marcan: tpw_rules is right, I created APKBUILD for my local testing and didn't pushed it to official alpine repo
<mps>
and I just told ashi that I have ready and don't need guide how to build mesa, to not waste time
chadmed_ has joined #asahi-alt
<mps>
and through years on different software projects no one told 'do not test published code'
jamespmo_ has joined #asahi-alt
jamespmorgan has quit [Ping timeout: 480 seconds]
kujeger has joined #asahi-alt
sven has joined #asahi-alt
cylm_ has joined #asahi-alt
cylm has quit [Ping timeout: 480 seconds]
chadmed_ has quit [Remote host closed the connection]
chadmed_ has joined #asahi-alt
jamespmorgan has joined #asahi-alt
jamespmo_ has quit [Ping timeout: 480 seconds]
<jannau>
mps: how often in that time did you "test" in random developer branches. I'd argue that random developer branches are not published code
<sven>
fwiw, 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.
<sven>
that 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"
<sven>
*though
<mps>
jannau: yes, I know. I never told that I push test I do to end users
<mps>
I merge to alpine only released things
<jannau>
that was a reply to "through years on different software projects no one told 'do not test published code'"
<mps>
sure. and I don't like to push to users even released things before I or someone else test them
<mps>
but for decades whenever I ask about test things not yet released usually devs says 'please test and report possible problems/bugs/...'
ashi has joined #asahi-alt
<sven>
"not released" is probably what's in asahi-edge
<mps>
sven: I think so, and I didn't pushed it to alpine
<sven>
what's beyond that is probably best described as "unfinished and possibly broken hacks shared between developers"
<chadmed_>
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
<chadmed_>
yeah that ^
<mps>
well, I still count asahi things as alpha
<chadmed_>
yeah so taking the pre-alpha stuff that everyone knows is broken and going "hey this is broken" is a waste of time
<mps>
though by some miracles they works fine for me ;)
<chadmed_>
not just the developers, but your own too
<j`ey>
not miracles, the asahi released stuff is the stuff that is meant to work!
<ashi>
Exactly. 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?
<ashi>
And your hacked dev branches, only allowed to be built by other core inner circle asahi devs. Yes, they should be private
<j`ey>
so you want it to be more closed? heh
<ashi>
It would be more honest.
<_jannau_>
you are free to test things and fix bugs but we are not interested in bug reports on random developer branches
<j`ey>
im not sure if random branches are dishonest..
<chadmed_>
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
<chadmed_>
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"
<_jannau_>
also fixing bugs in random branches is probably a bad idea since that easily could be just duplicate work
<chadmed_>
then try to build it and get all huffy and pissed when it doesnt work 100% immediately with no pain
<j`ey>
ashi: they've been honest about not wanting people to test (and then report issues) on random branches, seems like a fair and honest thing
<ashi>
chadmed_: 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
<mps>
j`ey: I understand that, and I never filled any bug because of this
<chadmed_>
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
<ashi>
j`ey: Yup, but publishing (ie. pushing) code to branches that are not intended to be built.... hmm...
<j`ey>
ashi: you are free to build it, but the point is to not report issues etc, if it doesnt work
<chadmed_>
the sticking point here
<ashi>
j`ey: okay, so when I said in #asahi-alt "I built it and it does not work" is that a bug report to devs?
<sven>
effectively yes.
<chadmed_>
yes this isnt the channel for it, and we know it probably doesnt build
<chadmed_>
again, refer to the mess that happened a couple of days ago
<chadmed_>
what a waste of everyones time
<j`ey>
I 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
<sven>
fair enough
<sven>
but usually the first quickly leads to the latter
<ashi>
j`ey: I was not asking for help
<j`ey>
ashi: yeah and this isn't all about what you've done!
<chadmed_>
no _you_ werent, and thats fine
<chadmed_>
but again, recall what happened a couple of days ago with that person who was "asking" for help
<chadmed_>
thats the situation we need to avoid
<ashi>
And 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.
<chadmed_>
and strongly discouraging the building of random non-canonical branches that are mostly feature incomplete is a good way to avoid it
<ashi>
I found the results impressive of your random branches
<ashi>
*knowing* that it might all fall apart
<ashi>
I 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.
<j`ey>
ashi: that wasn't 'enough', it wasnt just you, it has happened a lot, hence the rant
<ashi>
Okay so my takeaway is - "do whatever you want with the random branches but don't tell anyone". Does that fit?
<chadmed_>
more like dont broadcast it in these channels
<chadmed_>
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
<chadmed_>
then sook when they cant run witcher 3 with RTRT enabled
beeblebrox has joined #asahi-alt
chadmed_ has quit [Remote host closed the connection]
ashi has quit [Quit: Leaving]
r0ni has quit [Ping timeout: 480 seconds]
<alyssa>
chadmed: msaa is super broken on agx/next (-:
<alyssa>
literally agx/next right now is asahi/main plus a bunch of commits labelled as HACK, WIP, and/or known broken :-p