<drakonis>
slackware is 90s linux era personified, yeah?
<ldlework>
indeed
<gchristensen>
haha 100%
<drakonis>
a quick lane to the 90s
drakonis has quit [Ping timeout: 245 seconds]
drakonis_ has quit [Ping timeout: 264 seconds]
<elvishjerricco>
Switched to zfs 0.8 to see how much effect `zpool trim` would have on my 960 Pro... went from 84 MB/s to 738 MB/s sequential write.
drakonis_ has joined #nixos-chat
<gchristensen>
wat
drakonis has joined #nixos-chat
<elvishjerricco>
gchristensen: Yea, I'd always heard that SSD over-provisioning made trim not so necessary, but then I saw the performance improvements shown in the PR and decided to give it a go: https://github.com/zfsonlinux/zfs/pull/8419
<{^_^}>
zfsonlinux/zfs#8419 (by behlendorf, 27 weeks ago, merged): Add TRIM support
<elvishjerricco>
ZFS has probably written all over this disk many times by now, with how much crap I write to it temporarily. So if any systems would see a benefit, I'm not surprised it's ones like mine
drakonis_ has quit [Ping timeout: 264 seconds]
<pie_>
elvishjerricco: is that like a one time thing?
<pie_>
re: whole deal with enabling trim on luks etc
<pie_>
tbh my zf feels pretty luggish, its almost like im not running on an ssd. i *think* its zfs.
<elvishjerricco>
pie_: Trimming on ZFS over LUKS requires you to set `boot.initrd.luks.devices.<device>.allowDiscards = true;", and using ZFS 0.8
<pie_>
elvishjerricco: yeah but can i like do a one time bach trim or is that not how this works
<elvishjerricco>
Then to actually making trimming happen, you can either manually run `zpool trim` periodically to trim the whole disk, and/or you can enable autotrim on the pool to have it trim blocks as they're freed by the system (though it won't trim particularly small blocks like `zpool trim` will)
<elvishjerricco>
So it's probably best to enable autotrim, and every now and then run a full `zpool trim` (probably quite rarely, really)
<elvishjerricco>
though IIRC, trimming should have absolutely no effect on read performance. Probably want to double check me on that
<ldlework>
i feel like if you like factorio, you'd probably really find protein synthesis pretty fascinating
<makefu>
philipp[m]: can only be better than the current implementations (both in kernel and as fuse module). the kernel version resulted in kernel oops for me and the othe version is just pretty slow
drakonis has joined #nixos-chat
drakonis has quit [Read error: Connection reset by peer]
drakonis1 has joined #nixos-chat
drakonis_ has quit [Ping timeout: 245 seconds]
drakonis1 has quit [Read error: Connection reset by peer]
drakonis has joined #nixos-chat
freeman42x[NixOS has joined #nixos-chat
<jD91mZM2>
I just got an invite to the NixOS organization on GitHub. While it's definitely an honor... why did I get an invite to the NixOS organization?
<jD91mZM2>
etu: Yeah probably, thanks! I was wondering if it was that but I never left a comment on that thread (worried I won't have time to help much)
<etu>
jD91mZM2: yeah, but this is more or less everyone that is listed as a package maintainer
<jD91mZM2>
Oooh, nice
freeman42x[NixOS has quit [Remote host closed the connection]
freeman42x[NixOS has joined #nixos-chat
valebes has joined #nixos-chat
<samueldr>
sphalerite: not reproducibly
<sphalerite>
huh
<samueldr>
yeah...
drakonis has quit [Read error: Connection reset by peer]
drakonis has joined #nixos-chat
drakonis_ has joined #nixos-chat
drakonis has quit [Ping timeout: 252 seconds]
drakonis has joined #nixos-chat
drakonis_ has quit [Ping timeout: 264 seconds]
freeman42x[NixOS has quit [Quit: Leaving]
freeman42x[NixOS has joined #nixos-chat
endformationage has joined #nixos-chat
drakonis_ has joined #nixos-chat
drakonis has quit [Ping timeout: 252 seconds]
waleee-cl has joined #nixos-chat
drakonis has joined #nixos-chat
freeman42x[NixOS has quit [Quit: Leaving]
abbradar has joined #nixos-chat
averell has joined #nixos-chat
evanjs has quit [Quit: Configuring ZNC, sorry for the joins/quits!]
evanjs has joined #nixos-chat
<gchristensen>
hot dog I have come across a bassoon in almost definitely good shape in a price range I can afford :o
<__monty__>
You play the bassoon?
<samueldr>
(no and it's why the neighbours stole the last one)
<gchristensen>
lol
<__monty__>
Am I mistaken in thinking the bassoon's mostly just an orchestra instrument?
<gchristensen>
I played in school, but they're hopelessly too expensive for people to casually get and the market to buy them used is not good.
<gchristensen>
I bought one for $1,000 several years ago and it was unusable.
<__monty__>
o.O
<gchristensen>
yap.
<gchristensen>
so, having one come up on the market at a price which I can afford and in good shape is incredibly rare :)
<gchristensen>
incredibly rare that this is only the second time I've had the chance, and I've had requests out to music shops for ten years now
evanjs has quit [Ping timeout: 246 seconds]
<__monty__>
There's no professional players who'd be willing to part with an instrument they no longer use?
<gchristensen>
lots
<__monty__>
But not for reasonable prices?
<gchristensen>
but a good quality new bassoon is the same cost as a new steinway piano, and used bassoons are not much cheaper than new bassoons
<__monty__>
That's one expensive wooden tube.
<__monty__>
Are electronic bassoons a thing?
<gchristensen>
some people have put a guitar pickup on a bassoon
freeman42x[NixOS has joined #nixos-chat
vika_nezrimaya has joined #nixos-chat
<maralorn>
fwiw __monty__ I have just heard a bassoon presenting a nice piece without orchestra just two weeks ago.
valebes has quit [Ping timeout: 245 seconds]
<gchristensen>
what if nix intercepted build-time syscalls and started building all the packages at once irrespective of the build tree, and just halted any time a build is going to access a path which would be created by a different build also running
<jD91mZM2>
gchristensen: I may be out of the loop, but what's the problem being solved with that?
<averell>
sort of lazy building? i'm not sure it would result more than just recording which pkgs were never accessed though, and make that a warning or similar.
<emily>
jD91mZM2: faster builds, in theory
<emily>
you block less often
<gchristensen>
I'm just making stuff up, but yeah that is the sort of idea
<eyJhb>
"I'm just making stuff up" - gchristensen 2019
<gchristensen>
pretty much my entire career right there
<jD91mZM2>
Would be an interesting idea, but I'd love if it was possible to put it in a feature flag. ptrace can be disabled on lots of setups (IE containers), so it would fall back to always blocking on dependencies
<gchristensen>
it is almost certain to never happen
<emily>
I think there's a prototype build system or two based on FUSE or whatever that works like this
<eyJhb>
gchristensen: Just read it aloud for the guys on the project, +1 :p
<jD91mZM2>
Also the overhead of stepping through syscalls may not be worth it on machines like mine with no more than 4 cores.
<ashkitten>
if you can do hundreds of build processes at once, sure
<ashkitten>
it seems less helpful otherwise
drakonis has quit [Quit: WeeChat 2.5]
<emily>
you don't need to step through syscalls. you can just use FUSE
<emily>
or write a kernel module if you're up for it
<emily>
filesystem calls are allowed to block :)
<gchristensen>
heh
valebes has joined #nixos-chat
freeman42x[NixOS has quit [Remote host closed the connection]
<ashkitten>
that's honestly a pretty dang good idea
<hyperfekt>
gchristensen: That sounds sexy as hell
<ashkitten>
heh, so like, a FUSE filesystem for /nix/store that builds paths on demand when you try to access them
<ashkitten>
hmm, nix can't get to the derivation from just the store path tho can it? so it'd need to figure out what derivation to build >,,>
<infinisil>
It can
drakonis has joined #nixos-chat
<ashkitten>
oh, how?
<infinisil>
nix-store -q --deriver /nix/store/...
<infinisil>
is an interface to it
<infinisil>
It probably just does an sql query on the database
<ashkitten>
but is that only for built paths?
<infinisil>
Ah..
<infinisil>
Yeah probably
<joepie91>
gchristensen: the major obstacle I can see is stuff that doesn't cope well with fs latency
<gchristensen>
I love that y'all are actually thinking about it
<joepie91>
which is painfully much, based on my experience with sshfs mounts :P though maybe this is less of a problem for build tools
<joepie91>
sure, why not? the mad science is what I'm here for :D
<ashkitten>
gchristensen: i don't see how useful it'd be, but i don't see any reason not to consider it
<gchristensen>
it was just a throw-away comment realizing (pkgs.writeText "foo" "whatever ${bar}") doesn't need to wait for `bar` to exist :P
<joepie91>
the "wouldn't it be neat if... but surely that can't work" projects always turn out to be the best ones :P
<ashkitten>
how often are builds stuck waiting for only one package, though?
drakonis_ has quit [Ping timeout: 264 seconds]
<gchristensen>
no idea
valebes has quit [Quit: leaving]
<jD91mZM2>
"Science isn't about why, it's about why not!"
<ashkitten>
tldr, the baseline jit is much faster than the c++ bytecode interpreter but has a longer warmup time so they added another interpreter that shares caches and such with the baseline jit which helps with pages that contain Way Too Much Javascript
Myhlamaeus2 has joined #nixos-chat
evanjs has quit [Ping timeout: 258 seconds]
evanjs has joined #nixos-chat
<cransom>
pages with way too much javascript ... also known as 99% of the web.
Myhlamaeus2 has quit [Ping timeout: 268 seconds]
<vika_nezrimaya>
cransom: maybe less than 99% if the IndieWeb movement becomes more popular
<vika_nezrimaya>
we (IndieWeb users) try to keep JS to minimum; and try to make everythin curl-able, i.e. the main content remains viewable without JS.
<vika_nezrimaya>
(the curl-ablility requirement stems from usage of microformats2 semantic markup and several formats requiring machine-readable web pages, i.e. no JS to see main content)
<vika_nezrimaya>
If it's not curlable, it's not on the web!
hyperfekt has quit [Quit: hyperfekt]
<__monty__>
Sadly most sites are designed to make scraping *hard*. And curl's like the zeroeth level of scraping : )
<vika_nezrimaya>
__monty__: then make a difference yourself. Make your website scrapable! (if you have one; and if you don't, now is a good time to have one!)
<__monty__>
I sympathize with the goals. Just sharing my sadness ; )
hyperfekt has joined #nixos-chat
<vika_nezrimaya>
I want the Web to have open APIs. Even if it's REST snowflake APIs, it's at least an open API. I wish taxi service I use and my bank had open APIs
Myhlamaeus2 has joined #nixos-chat
<vika_nezrimaya>
I imagine how a bank may work. You make a cryptographic key, and then ask banks to issue checks in your name, just like old times
<vika_nezrimaya>
except an issue statement is signed by your key, and a check is a token
<vika_nezrimaya>
a token, that the receiver can redeem for money
<vika_nezrimaya>
I hope I'm reinventing the wheel
<vika_nezrimaya>
Please tell me someone already invented this!!! >.<
<vika_nezrimaya>
we have cryptocurrency though, but it's not well-accepted in most countries
<emily>
it's called bitc- yes
<vika_nezrimaya>
and some countries even want to prohibit it
<emily>
people mostly don't want cryptographically-inviolable finance systems
<vika_nezrimaya>
maybe because they want to cheat them?
<emily>
having courts to sort out ambiguity and unfairness in squishy cases where there isn't a clear hard-and-fast answer is generally considered an advantage
<cransom>
we have users that complain about accessability (which, if a web site is curlable, the screen reader would be fine) and on the other end we have managers who fancy them selves as being developers, despite having not coded in the company code base in years, saying REACT. MAKE IT FANCY. ALL THE JS.
<emily>
I mean, have you ever done a chargeback? :)
<tokudan>
vika_nezrimaya, there is a public standard for online banking named FinTS. pretty much every bank in germany supports it. so if there is no standard in your country, that's not because there is no standard, it's because your banks are lazy
<maralorn>
tokudan: I am not sure if the adjective lazy is applicable.
<cransom>
greedy?
<cransom>
thats usually applicable
<tokudan>
lazy works, i think. greedy isn't exact enough, in my opinion. the banks could offer better customer service, but are too lazy to do it
<tokudan>
also transferring money via checks is strange... why not just give your bank the information to transfer X amount of money to account number Y at bank Z? no possibility of checks getting lost in the post
evanjs has quit [Quit: Configuring ZNC, sorry for the joins/quits!]
evanjs has joined #nixos-chat
andi- has quit [Remote host closed the connection]
andi- has joined #nixos-chat
<gchristensen>
is it possible to have obs continue recording a window which is not in the forefront?
<drakonis_>
yes
<drakonis_>
if you focus specifically on the window
<drakonis_>
i think
<infinisil>
Is there some decent website for sharing binary files?
<infinisil>
I guess gist would work actually, even if it's totally not text
<infinisil>
Actually no it doesn't, gist doesn't allow non-utf8
<__monty__>
ashkitten: Wouldn't nix *already* know about the derivation? It just wouldn't know *which* of its inputs generates a path. But that doesn't matter, it could just build everything in parallel and block on paths not being in the database.
<ashkitten>
__monty__: yeah that's fair, but we'd have to ensure that for every referenced store path there is a builder running
qyliss^work has quit [Quit: bye]
<ashkitten>
if a build is stuck waiting for a path that never appears, that can be a problem
<gchristensen>
it'd have to block on the /nix/store/<hash>-<nam> not the sub-path
<__monty__>
True, though that's fairly easily solved by queueing deps first, recursively.
<__monty__>
gchristensen: Hmm, why exactly?
<ashkitten>
i'm guessing because the store path might appear but not contain the sub-path
<ashkitten>
which then blocks the build indefinitely
<gchristensen>
yeah
<__monty__>
And it should then fail with an error about a missing path, rather then not finish?
<ashkitten>
we can have the filesystem talk to the nix daemon and check what builds are running to ensure that all queued paths are actually being built, and return ENOENT on paths that appear without the sub-path or if there is no builder for them
<ashkitten>
so you have to start dependency builders before dependent ones, but that's not an issue
valebes has quit [Read error: Connection reset by peer]
valebes has joined #nixos-chat
<ashkitten>
i kinda wanna try doing this but i haven't even touched nix itself yet
<ashkitten>
and that's scary lol
* __monty__
whisper "Do iiiit." from a dark corner of ashkitten's room
<ashkitten>
lol
<ashkitten>
why are you under my girlfriend's desk
<ashkitten>
do you want a cookie
<samueldr>
ashkitten: every time I've had to look or touch nix's code I was pleased to see a maintainable and approachable codebase
<samueldr>
ashkitten: what I mean is: do it
<ashkitten>
samueldr: ahahah, i'll check it out
<ashkitten>
can i test custom nix in a container?
<__monty__>
ashkitten: Is that a leading question? -_-
<ashkitten>
__monty__: no, i want to give you a cookie if you want one
<ashkitten>
a good thing with this custom filesystem adapter approach is that you can just enable it on top of any existing filesystem without disturbing it
<ashkitten>
so you don't have to change anything about the nix store if you enable this path based parallelization approach
<ashkitten>
it'll just be an overlay
<__monty__>
Hmm, is this starting to sound a bit too good to be true? 🤔
waleee-cl has quit [Quit: Connection closed for inactivity]
valebes has quit [Read error: Connection reset by peer]
valebes has joined #nixos-chat
drakonis_ has joined #nixos-chat
drakonis has quit [Ping timeout: 264 seconds]
drakonis has joined #nixos-chat
valebes has quit [Read error: Connection reset by peer]
valebes has joined #nixos-chat
valebes has quit [Read error: Connection reset by peer]
ddima has joined #nixos-chat
valebes has joined #nixos-chat
valebes has quit [Read error: Connection reset by peer]
valebes has joined #nixos-chat
__monty__ has quit [Quit: leaving]
psyanticy has quit [Quit: Connection closed for inactivity]
valebes has quit [Ping timeout: 268 seconds]
valebes has joined #nixos-chat
qyliss has joined #nixos-chat
freeman42x[NixOS has quit [Ping timeout: 258 seconds]
evanjs has quit [Killed (hitchcock.freenode.net (Nickname regained by services))]
evanjs has joined #nixos-chat
<evanjs>
jD91mZM2: nope, you’re required to make more rust nix stuff now that you’re on the team 😝
<hyperfekt>
samueldr: How does this compare to niv? It seems like updating sources is a thing both in nixpkgs and in configurations and I wonder whether the same tool can be used for both.
<samueldr>
I have no idea what niv is
<samueldr>
I just saw that while updating another package
<hyperfekt>
tool to update source hashes and versions. seems there's a lot of overlap, but this one seems geared at nixpkgs and the other at individual projects https://github.com/nmattia/niv