Email or username:

Password:

Forgot your password?
Haelwenn /элвэн/ :triskell: 's posts Post Back to profile
Haelwenn /элвэн/ :triskell:

ERROR: Unsupported linker, only bfd, gold, and lld are supported, not ld.bfd.

Fucking buildsystems (meson this time), none of them are good.

18 Apr 2022 at 6:16 | Open on queer.hacktivis.me
21 comments
Haelwenn /элвэн/ :triskell:
Consider:
- Be as agnostic as possible: Don't hardcode, use pkg-config or environment variables on pretty much everything
- Don't test if the OS works, that's not our problem
- Strongly avoid workarounds, OS should be fixed instead or more portable methods should be used
18 Apr 2022 at 6:25 | Open on queer.hacktivis.me
p
@lanodan Mostly seconded, except:

> Don't hardcode, use pkg-config or environment variables on pretty much everything

Build systems that are normal makefiles with a 10-line "This works on Linux x86" and "This works on OpenBSD on ARM" that you call with `make CONF=linux386` or whatever, those tend to be easier to debug than some flaky pkg-config problem in a 300k generated makefile.
@lanodan Mostly seconded, except:

> Don't hardcode, use pkg-config or environment variables on pretty much everything

18 Apr 2022 at 6:36 | Open on freespeechextremist.com
Haelwenn /элвэн/ :triskell:
@p
> This works on Linux x86

Except Linux does not exists as an OS, it's at least often easy to fix those things with few sed commands.

> generated makefile

That's also a thing which IMHO shouldn't be done, do it like musl and drop a config.mk
@p
> This works on Linux x86

Except Linux does not exists as an OS, it's at least often easy to fix those things with few sed commands.
18 Apr 2022 at 6:53 | Open on queer.hacktivis.me
Johnny Retard ඞ
@lanodan @p
> Except Linux does not exists as an OS
😩 why did bsd have to be fucked with that legal battle :(
18 Apr 2022 at 23:31 | Open on bae.st
p
@MischievousTomato @lanodan We'd be asking that about Linux if BSD took off. Linux is terrible because it took off and people got obsessed with making it do everything for everyone so it would kill Windows.

:dmr: "Unix has retarded operating systems research by ten years, Linux by another twenty."
19 Apr 2022 at 12:21 | Open on freespeechextremist.com
Shebang! :blobpeektrans:
@p @MischievousTomato @lanodan His quote really isn't fair, he's being salty here. You want to blame OS research point the finger at Microsoft and Apple, the dominant forces here. They've worked to make everything so much worse intentionally, where Linux its flaws are at least accidental.
19 Apr 2022 at 12:23 | Open on freespeech.group
p
@shebang @MischievousTomato @lanodan

> where Linux its flaws are at least accidental.

That is the problem, it's not the flaws, it's because it's a platform that works well enough for people to standardize. MS/Apple are making OS products but they're not running big research projects: just enough research to improve their products and to attract the big-brain PhDs. Linux is simple and widespread and complies more or less with POSIX, a spec based on 1980s Unix intended to get vendors to standardize. The point of the quote is that Unix (which he co-created) was good enough to kill research, people just mashed Unix in without thinking, and it's worse with Linux.
@shebang @MischievousTomato @lanodan

> where Linux its flaws are at least accidental.

That is the problem, it's not the flaws, it's because it's a platform that works well enough for people to standardize. MS/Apple are making OS products but they're not running big research projects: just enough...
19 Apr 2022 at 15:29 | Open on freespeechextremist.com
Haelwenn /элвэн/ :triskell:
@p @MischievousTomato @shebang I think the true worst is seeing all those Unix-clones in things like osdev.org just repeating the same damn mistakes.

Not only is UNIX dead, it’s starting to smell really bad.
— Rob Pike circa 1991


And I would argue that Plan9, Inferno and BeOS also are dead, they just don't really stink yet.
@p @MischievousTomato @shebang I think the true worst is seeing all those Unix-clones in things like osdev.org just repeating the same damn mistakes.

Not only is UNIX dead, it’s starting to smell really bad.
19 Apr 2022 at 17:30 | Open on queer.hacktivis.me
p
@lanodan @MischievousTomato @shebang I'll take a dead system over an epileptic one, but I don't think any of them are dead. Linux-the-ecosystem is spastic and it's adorable except when you have to attempt to get something done.
spaz.gif
spaz.gif
19 Apr 2022 at 19:08 | Open on freespeechextremist.com
Haelwenn /элвэн/ :triskell:
@p @MischievousTomato @shebang Well, if I would seriously take Plan9 (the most likely one), I would very likely end up seriously modifying it to the point where it's not a Plan9 anymore.
(and no, not in a Unix/Linux direction, I want that gone)
19 Apr 2022 at 19:10 | Open on queer.hacktivis.me
p replied to Haelwenn /элвэн/ :triskell:
@lanodan @MischievousTomato @shebang You might be surprised; the more I did with Linux, the more I wanted to change, but the opposite turned out with Plan 9/Inferno. I mean, I'm still patching things but it's little stuff, like adding '-c' to Inferno's uniq(1), changing a handful of minor behavioral things in acme, things like that. I wrote a tiny Redis client in Limbo, I wrote a whois client. The systems are really nice as-is, I spend no time wishing I didn't have to hack around a bad decision, I just do my normal activities, writing small scripts to build up my environment so that the things I want to do are convenient. Like this, I want to see what I have changed before saving a file sometimes, or I change it in one session while it's still in an editing buffer on another, little usability things:

#!/bin/rc
adr = /mnt/acme/$winid
ofn = `{sed 's/ Del.*//' $adr/tag}
diff $ofn $adr/body && echo clean > $adr/ctl

Small stuff like that. It's like teaching the system what I want to do.

I'm interested in where you want an OS to go, though I asked that indirectly in the other post, like what stuff you want to take from TempleOS. Like, what I want is a system that I can hack as needed, flexible enough to let me hack the bits I want to hack, comfortable for hacking (itself and the things that I'm writing), quiet so that I don't have to stop it doing anything that I don't want it to do, and able to tie together other machines so that I can play with the whole system, so Plan 9 and Inferno both fit great. It's not sarcasm or rhetorical, I'm not trying to advocate, I'm interested in the question in a really literal sense: what do you want from an OS?
@lanodan @MischievousTomato @shebang You might be surprised; the more I did with Linux, the more I wanted to change, but the opposite turned out with Plan 9/Inferno. I mean, I'm still patching things but it's little stuff, like adding '-c' to Inferno's uniq(1), changing a handful of minor behavioral things in acme, things like that. I wrote a tiny Redis client in Limbo, I wrote a whois client. The systems are really nice as-is, I spend no time wishing I didn't have to hack around a bad decision,...
19 Apr 2022 at 19:46 | Open on freespeechextremist.com
Haelwenn /элвэн/ :triskell: replied to p
@p @MischievousTomato @shebang
Well to me the annoying parts of Plan9 (listing the good parts would be too long):
- Lack of things like keyboard accessibility, using Plan9 as a daily driver would probably damage my wrists further, I can still use a mouse fine in most situations just not *that* much
- No multi-monitor support (not that there is good GPU support anyway but I don't care much about that except for efficient video decoding)

And then there is few autisms:
- Move more things out of the kernel, as I think that would make it easier to create and maintain drivers while also getting better stability (haven't ran Plan9 enough to judge that part though)
- Probably write another file storage, hjfs and cwfs have been unreliable in my experience

Meanwhile the annoying parts of most Unix-likes:
- (Linux) I want a base system dammit, but it doesn't means I want a monorepo with vendored tools
- Multi-architecture support *sucks* (specially cross-compiling)
- The kernels are way too monolithic, wtf are things like the filesystems doing in it
- No one has a VT200 on their desk except for retro-nostalgia purposes
- The filesystem hierarchy is still stuck to early-Unix and the mob doesn't knows how to fix it, they just add more symlinks
- They more have to do with the ass-backward stuff from GNU than actual UNIX®
@p @MischievousTomato @shebang
Well to me the annoying parts of Plan9 (listing the good parts would be too long):
- Lack of things like keyboard accessibility, using Plan9 as a daily driver would probably damage my wrists further, I can still use a mouse fine in most situations just not *that* much
19 Apr 2022 at 20:21 | Open on queer.hacktivis.me
p replied to Haelwenn /элвэн/ :triskell:
@lanodan @MischievousTomato @shebang

> - Lack of things like keyboard accessibility

Yeah, you can tweak this kind of thing; sam is pretty keyboard-friendly out of the box. You can script keyboard events arbitrarily, even, just put a pipe between the keyboard device and rio. I don't because it's not useful but I used to do this kind of thing with Inferno, but I can understand the wrist issues with mice. (Although Thinkpads are nice for this, but lateral stress on my index finger tends to make the joints sore if I use just my laptop for a few days.)

> No multi-monitor support

Don't quote me on this, but I think that's been fixed. I usually use drawterm anyway; there's stuff that lives in Linux that I can't avoid. (Linux is my new Windows.) You can run $n operating systems now that you can get a pile of computers without even having to dumpster-dive and you can run 30 VMs on each of them.

> Move more things out of the kernel, as I think that would make it easier to create and maintain drivers while also getting better stability (haven't ran Plan9 enough to judge that part though)

Yeah, it doesn't work like that; it's trivial to play with drivers, and it's ridiculously stable but it doesn't actually need to be: your files live on the fileserver (which should be stable), you run your stuff on a CPU server and connect to a terminal server and you can reboot basically any of those without consequence, and it's had kexec() since forever (reboot $kernelpath) so as long as you don't write a kernel that crashes the machine and also hoses remote debugging, it's like a few seconds to test a new kernel. So stability's less of an issue, but I've never gotten it to crash by normal means.

But I'm interested in what you want, like analogous to the stuff I was talking about. What do you want your OS to do? Like, one step up from feature bullet points.
@lanodan @MischievousTomato @shebang

> - Lack of things like keyboard accessibility

Yeah, you can tweak this kind of thing; sam is pretty keyboard-friendly out of the box. You can script keyboard events arbitrarily, even, just put a pipe between the keyboard device and rio. I don't because it's not...
19 Apr 2022 at 22:04 | Open on freespeechextremist.com
Haelwenn /элвэн/ :triskell: replied to p
@p @MischievousTomato @shebang A hacky way of getting multimonitor on Plan9 would be something like grabbing some SunRay or using my dusty ARM SBCs as thin clients, heck, why not a mix just because I can.
(I hate ARM shit though, I want a boot ROM, even if it's a cursed BIOS)

I think it's stuck to not having multimonitor for me because all my GPUs are AMD and I guess it would only support that on Intel ones where it already has more than just VESA.

Thing I wish to be able to do with Plan9 are:
Replace at least most of the terminal/GUI stuff that I have on linux.
I'm tired of things in curses that never speak/listen to other programs and are stuck to a crude interface. Like shit today means I need at least a basic HTML viewer for RSS/Atom, parts of Email, … and while you can emulate a terminal in curses apps to spawn lynx/w3m in them it's a nightmare to actually use since shortcuts conflict between each others in an unpredictable manner. And using --dump is just a hack and it's forever stuck to plain text.
Meanwhile Qt/GTK/… makes me want to throw everything away because one is just proprietary-developed + open-source released, while the other is just happily jamming into a wall everytime the CADT refreshes itself.

And I do wish to move my audio programs to Plan9, I could also move it to a BSD but… I don't really like those systems, they're not as hackable as I want an OS to be.
Pipewire on the linux side of things at least seems to be going in a good enough direction by linux standards but it's not going to beat just simply writing into /dev/audio and easily being able to hack on it with knowing that the rug won't be pulled from under you at some point by the CADT mob making yet another half-baked solution be the de-facto standard.
@p @MischievousTomato @shebang A hacky way of getting multimonitor on Plan9 would be something like grabbing some SunRay or using my dusty ARM SBCs as thin clients, heck, why not a mix just because I can.
(I hate ARM shit though, I want a boot ROM, even if it's a cursed BIOS)
19 Apr 2022 at 22:32 | Open on queer.hacktivis.me
p replied to Haelwenn /элвэн/ :triskell:
@lanodan @MischievousTomato @shebang

> I think it's stuck to not having multimonitor

Yeah, I mean, that's less relevant if you're just using drawterm to talk to a headless box or a VM.

> Replace at least most of the terminal/GUI stuff that I have on linux.

So, my experience is some stuff you can't really replace so you have to talk to Linux, some stuff is better and I don't miss the original thing at all, some of it's the same. It's like using a different programming language and nowadays it's about as easy to use 50 OSs as it is to use 50 languages.

> Meanwhile Qt/GTK/… makes me want to throw everything away

Seconded. My experience with GUI stuff prior to Inferno was all gtk/qt/etc. with some occasional "I guess I have to figure out how to make Windows have a systray icon for this thing" and I hated all of it. Inferno has a port of Tk so I learned Tk and it made me wonder why I hadn't sooner. Like you can bash out a GUI in wish in 30 seconds, it's a pipe, you can make a GUI in awk. (The temperature monitor from that screenshot, that's an awk script.)

> And I do wish to move my audio programs to Plan9

Oh, it's pretty easy to do. I like cmus, personally, but Plan 9's pretty easy to do sound stuff for. (Have you seen this? https://git.sr.ht/~ft/neindaw )
@lanodan @MischievousTomato @shebang

> I think it's stuck to not having multimonitor

Yeah, I mean, that's less relevant if you're just using drawterm to talk to a headless box or a VM.
20 Apr 2022 at 0:17 | Open on freespeechextremist.com
Haelwenn /элвэн/ :triskell: replied to p
@p @MischievousTomato @shebang
> Tcl/Tk

I'm not sure if I gave it an honest try once, but I don't feel it at all.
Also their fucking source code browser just took me for a fucking bot while I was going around at a rather slow pace. (pic attached)

> neindaw

I haven't tried using it yet but I don't into DAWs or even livecoding so it having an interface makes me take it as "Oh, yet another thing I won't use".
I want to program it entirely, even before I got in computers I wished I could just give the music sheets directly to my harp rather than have to practice everyday.
It kills what I find fun in music.
screen.png
@p @MischievousTomato @shebang
> Tcl/Tk

I'm not sure if I gave it an honest try once, but I don't feel it at all.
screen.png
20 Apr 2022 at 0:49 | Open on queer.hacktivis.me
:umu: :umu:
@lanodan

>- Don't test if the OS works, that's not our problem

this. Autismtools is a relic of time when it was really needed. Now it's better to fix OS itself, rather than put Yet Another Workaround for specific OS.

Though I still check for C standards, extensions and ofc if pkg-config delivered working library, it helped to debug problems early.
@lanodan

>- Don't test if the OS works, that's not our problem

this. Autismtools is a relic of time when it was really needed. Now it's better to fix OS itself, rather than put Yet Another Workaround for specific OS.
18 Apr 2022 at 17:43 | Open on expired.mentality.rip
Go Up