channel logs for 2004 - 2010 are archived at http://tunes.org/~nef/logs/old/ ·· can't be searched
#osdev2 = #osdev @ Libera from 23may2021 to present
#osdev @ OPN/FreeNode from 3apr2001 to 23may2021
all other channels are on OPN/FreeNode from 2004 to present
http://bespin.org/~qz/search/?view=1&c=osdev2&y=24&m=7&d=10
00:59:00 <kazinsal> zoinks it's hot out
00:59:00 <kazinsal> air conditioning is struggling to keep up
01:10:00 <heat> i really don't want to write a gc :(
01:27:00 <geist> kazinsal: yah no kidding, looks like it's a bit hotter up there
01:37:00 <kazinsal> yeah
01:38:00 <kazinsal> topped out at 93 F today with a feels like of 99
13:03:00 <GeDaMo> https://www.theregister.com/2024/07/10/dshr_on_news_vs_x/
13:03:00 <bslsk05> www.theregister.com: Lead developer of original X rival NeWS looks at why it died • The Register
14:22:00 <nikolapdp> did NickServ just completely die
14:24:00 <GeDaMo> It just responded to me
14:24:00 <nikolapdp> yeah it took a while but it responded to me too
17:16:00 <heat> xterm cross compile looks fucked
17:16:00 <heat> it picks up on host dependencies...
18:17:00 <Ermine> why would you cross compile xterm
18:17:00 <heat> onyx
18:18:00 <heat> meanwhile i tried running the GNU hello configure script on onyx and i think i'm leaking anonymous memory
18:18:00 <heat> it might be time for the ultimate linux vm larp, clean that shit all up
18:18:00 <Ermine> it's X application, since when do we have X11 there
18:18:00 <Ermine> xterm is
18:18:00 <heat> i have X11 ports locally
18:18:00 <nikolapdp> oh did you get drm working
18:18:00 <Ermine> WOAH
18:19:00 <heat> no, this one's running on fbdev still
18:19:00 <nikolapdp> ah
18:20:00 <Ermine> seems like there's no MINIMAL terminals which aren't naziware and don't require opengl
18:21:00 <heat> xterm doesn't require opengl
18:21:00 <Ermine> besides xterm
18:21:00 <nikolapdp> it's also not minial exactly
18:21:00 <nikolapdp> is it
18:21:00 <heat> xterm is more minimal than konsole or gnome terminal or whatever
18:21:00 <nikolapdp> that is true
18:22:00 <Ermine> xterm makes my eyes weep
18:22:00 <GeDaMo> Are there terminal emulators that only use the framebuffer?
18:22:00 <Ermine> seems like no
18:23:00 <nikolapdp> i am sure there are some
18:23:00 <GeDaMo> https://wiki.archlinux.org/title/KMSCON ?
18:23:00 <bslsk05> wiki.archlinux.org: KMSCON - ArchWiki
18:23:00 <Ermine> > KMS
18:23:00 <Ermine> it's linux thing
18:24:00 <Ermine> ah, it's not a requirement actually
18:24:00 <Ermine> neither it's x11
18:24:00 <heat> developed by the most mentally sane linux developer
18:24:00 <nortti> there's also https://github.com/aligrudi/fbpad
18:24:00 <bslsk05> aligrudi/fbpad - A small Linux framebuffer virtual terminal (14 forks/104 stargazers)
18:24:00 <Ermine> then it became systemd-consoled
18:25:00 <GeDaMo> "I found a bunch of framebuffer terminal emulators, including fbcon, fbterm, bterm, yaft, and fbpad." https://bylr.info/articles/2022/10/29/til-textual-fb-term/
18:25:00 <bslsk05> bylr.info: TIL: running Textual on a framebuffer terminal emulator for Linux | bylr.info
18:27:00 <heat> i can only use terminal emulators with glass effect support
18:27:00 <nikolapdp> that disqualifies xterm though
18:27:00 <heat> i know
18:27:00 <heat> i only use BLOATED software like konsole and systemd
18:27:00 <heat> and glibc
18:28:00 <GeDaMo> And onyx? :P
18:28:00 <heat> somewhat
18:28:00 <heat> actually did some memory tuning a month ago or so, made it use significantly less memory at boot
18:29:00 <heat> it was all about tuning slab sizes and making sure i dont e.g touch a 4M-sized kmalloc cache for a one time boot alloc
18:30:00 <heat> it still takes some good memory at boot but the system's large anyway
18:30:00 <Ermine> konsole is fine imo
18:31:00 <Ermine> it comes with some kind of ssh manager but I didn't figure it out yet
18:31:00 <heat> boot to shell takes 20M on my end
18:32:00 <Ermine> -12MB
18:32:00 <Ermine> Great job
18:32:00 <Ermine> enough to fit into my old dir-300 router which only has 32MB of ram
18:33:00 <GeDaMo> For a second I thought that 20M meant minutes :|
18:33:00 <heat> hah no it's mostly instantaneous
18:33:00 <nikolapdp> i managed to boot artix with the default kernel to only use about 40mb of ram with a full x session
18:33:00 <nikolapdp> the kernels have ballooned since though
18:34:00 <nikolapdp> so not sure if that's possible now
18:34:00 <Ermine> take 2.4 kernel and you'll be ok
18:34:00 <heat> except with openrc, openrc with no interesting networking services actively slows down the boot process for 2-3s
18:34:00 <nikolapdp> heat, i'd be worried if it wasn't instant
18:34:00 <nikolapdp> Ermine it was in the 4.x series
18:35:00 <Ermine> vendor's patch for that dir-300 kernel is 30KLOC
18:35:00 <heat> nikolapdp, booting tends to be !instant
18:36:00 <nikolapdp> depending on what you do i guess
18:36:00 <nikolapdp> i assume you aren't enumerating all of physical memory at boot
18:36:00 <heat> sure i am?
18:36:00 <nikolapdp> you aren't doing the deferred thing that linux does
18:36:00 <heat> i'm not
18:36:00 <Ermine> who doesn't enumerate physical memory
18:37:00 <heat> the deferred linux struct page init thing isn't done by default AFAIK
18:37:00 <nikolapdp> right, don't know that part
18:37:00 <nikolapdp> i guess it's almost instant if you don't need to enumerate terrabytes of ram
18:37:00 <heat> and/or you don't have hard realtime requirements
18:38:00 <Ermine> on systems with such amounts of ram you don't care about os boot times
18:38:00 <heat> you defo do
18:38:00 <Ermine> since POST is already slow af
18:38:00 <heat> there was a fun slideshow somewhere of intel doing micro-optimization for a SoC that was going into a car
18:38:00 <heat> basically they needed boot to be like 1-2s max
18:38:00 <nikolapdp> do SoCs have terrabytes of ram
18:39:00 <heat> no
18:39:00 <nikolapdp> i assume that's what Ermine was referring to
18:39:00 <heat> a lot of it was mem init deferral, async driver probing, i915 quirks
18:39:00 <Ermine> i think those hp servers reached bootloader in 2 or 3 minutes
18:39:00 <netbsduser> apple apparently rejected making boot times fast, in favour of making good use of standby and suspend to disk
18:39:00 <mjg> dude
18:39:00 <Ermine> and that was 64Gb of RAM
18:39:00 <mjg> freebsd was majorly sped up in terms boot
18:39:00 <heat> PESSIMAL MAN IS HERE
18:39:00 <heat> teach me senpai
18:39:00 <mjg> most of it was literally "prod this device, now fucking sleep for a literal second"
18:39:00 <mjg> or rather busy wait
18:40:00 <mjg> tons of legacy shit which either did not need to be there or could have been deferred
18:40:00 <GeDaMo> I remember a trick, possibly on old Macs where they would show a saved image of the desktop while booting before the system was ready for the user
18:40:00 <heat> linux doesn't do async driver init by default for garbage compat reasons
18:40:00 <heat> you need to opt in
18:40:00 <nikolapdp> GARBAG
18:42:00 <Ermine> how much time can linux save if it did async init and deferred struct page init?
18:42:00 <heat> https://lpc.events/event/4/contributions/281/attachments/216/617/LPC_2019_kernel_fastboot_on_the_way.pdf
18:42:00 <GeDaMo> optimal in, pessimal out
18:42:00 <heat> there it is
18:42:00 <heat> from 3s to 300ms
18:42:00 <bslsk05> lpc.events <no title>
18:42:00 <Ermine> noice
18:43:00 <mjg> GeDaMo: now that does not surprise me tho
18:43:00 <mjg> there is a classic: progress bar goes forward while nothing is happening
18:43:00 <mjg> gmail does it for example
18:43:00 <mjg> try it out -- cut your net connection and refresh
18:43:00 <GeDaMo> Knowing that the system is still working is nice :P
18:44:00 <mjg> some ui folk elsewhere told me it's all about "user experience"
18:44:00 <nikolapdp> gmail dropped the html only ui
18:44:00 <nikolapdp> sadge
18:44:00 <mjg> i feel it
18:44:00 <Ermine> nikolapdp: use mail clients
18:44:00 <nikolapdp> yeah i do, i just didn't for gmail because i don't need it that often
20:03:00 <Ermine> arch fonts config is not great
20:05:00 <nikolapdp> is it
22:17:00 <heat> i dont get how im leaking the fucking memory :(
22:29:00 <heat> i found out how i was leaking most of the memory, though there's still a big leak somewhere
22:30:00 <mjg> no leaks in rust
22:30:00 <mjg> wait
22:31:00 <heat> no leaks in C++ too, but i'm an idiot that thinks he's writing C
22:32:00 <heat> SUPER OPTIMIZED RADIX TREE WHERE THINGS ARE UNSIGNED LONGS isn't too compatible with RAII or anything similar
22:32:00 <mjg> can you please not talk about "super optimized" and onyx
22:33:00 <heat> cry
22:33:00 <heat> i'm faster than your favourite BSDs
22:33:00 <mjg> i am doing it, yes
22:34:00 <mjg> maybe in a speedrun to crash
22:34:00 <mjg> unless you think obsd is my fav
22:34:00 <heat> i'm faster than net
22:34:00 <mjg> that is expected
22:34:00 <mjg> it's not hard to be faster than free either for the most part
22:34:00 <mjg> fucken damage all over
22:35:00 <mjg> but then again, one pokes at linux...
22:35:00 <mjg> does gcc run under onyx?
22:35:00 <mjg> i happen to have a nasty bench
22:36:00 <heat> yes
22:36:00 <mjg> https://dpaste.com/6LSSDLBJD
22:36:00 <bslsk05> dpaste.com <no title>
22:36:00 <mjg> plop a hello world into /tmp/src.c
22:36:00 <heat> i'm debugging a configure script going poopy and leaking memory all over
22:37:00 <mjg> q is how f45t and does it crash if you -j 8
22:37:00 <mjg> or whatever coren you got
22:37:00 <mjg> threaden
22:37:00 <heat> why is that nasty?
22:37:00 <mjg> it wont crash onyx?
22:37:00 <heat> it looks like a normal, eh, uh, configure script
22:37:00 <mjg> run it in parallel
22:37:00 <mjg> 's my thing here
22:43:00 <heat> i'll consider it mon
22:43:00 <heat> i'm dealing with a more important thing here
22:43:00 <mjg> twas a side remark
22:46:00 <heat> i suspect i'm only leaking actual anon memory here, not MAP_PRIVATE anon
22:47:00 <heat> if i create a new branch called vm2 will oracle sue me
22:48:00 <mjg> ask jls for permission
22:48:00 <kazinsal> they'll extrajudicially sentence you to orbital bombardment, but instead of tungsten rods they just drop a lawyer on your house
22:49:00 <heat> mjg, ironically i'm waiting for that mofer's book before COPYING LITERALLY ALL IDEAS
22:51:00 <mjg> lol
22:51:00 <mjg> n/c
22:51:00 <mjg> does he know you refer to him as 'mofer'
22:51:00 <mjg> :X
22:52:00 <mjg> kind of rude innit
22:52:00 <heat> you are right i did overstep my boundaries
22:52:00 <heat> mofer is for mjg and ermine only
22:52:00 <mjg> indeed, unbecoming for that network
22:52:00 <mjg> it is not for mjguzik thouhg
22:53:00 <heat> OH i've been speaking with a mjguzik fella sometimes
22:53:00 <heat> kinda weird he has the same name as you
22:53:00 <heat> but he does not say mofer so i'm sure it isn't you
22:53:00 <mjg> sounds like dude has a broomstick up his ass
23:07:00 <kof673> so....ye olde winelf toolchain and loader works, notwithstanding a "proc" "prog" typo in an elf struct lol > wine ~/gelfload_mingw32/bin/gelfload-i386.exe obj_winelf_gcc_4_4_2/tmaloc.elf so...yeah....
23:12:00 <heat> https://gist.github.com/heatd/33c658fc1ac03da3ba0aa51ca1724483 something's sussy
23:12:00 <bslsk05> gist.github.com: gist:33c658fc1ac03da3ba0aa51ca1724483 · GitHub
23:12:00 <gog> hi