gchristensen changed the topic of #nixos-chat to: NixOS but much less topical || https://logs.nix.samueldr.com/nixos-chat
infinisil has quit [Quit: Configuring ZNC, sorry for the joins/quits!]
infinisil has joined #nixos-chat
waleee-cl has quit [Quit: Connection closed for inactivity]
infinisil has quit [Quit: Configuring ZNC, sorry for the joins/quits!]
infinisil has joined #nixos-chat
<Gilfoyle-> So distri looks interesting
<Gilfoyle-> Guess we have some competition finally
<drakonis> guix is also competition i guess
<infinisil> Gilfoyle-: Well distri doesn't seem to focus on reproducibility
<drakonis> it has less trouble running third party software
<andi-> Distri also still requires the maintainer to bump meta package versions to trigger rebuilds. So ABI compat is the same issue as with Debian etc..
<gchristensen> guh
<andi-> After his talk at the GPN in Karlsruhe I talked to him about that. He didn't get why NixOS has "these weird hashes in the name of packages".
<drakonis> he's an ex debian maintainer, isn't he?
<andi-> He said he could just write tooling for that.. A libc change would require a bump of almost all meta versions in the repo. Also I think the nix-env benchmark misses the point. Nix also has this walk the Dependencies tree download/realise approach. Naturally downloads that have dependencies between each other aren't being done in parallel. Could be fixed ofc...
<andi-> drakonis: yeah, one of the few that wanted to improve things
<andi-> Also why am I up this early ? /o\ Sleep schedule nonexistent
<Gilfoyle-> Hmm, alright powerdns and the sqlite backend are a pain
<samueldr> you're up early? I assumed you were up late
<drakonis> its 1am my time
<Gilfoyle-> Okay how is this dang db ro?
<Gilfoyle-> It's got full write privs on it.
<andi-> samueldr: yeah early.. I got to bed 4h ago and now I am awake..
drakonis has quit [Ping timeout: 250 seconds]
drakonis has joined #nixos-chat
endformationage has quit [Ping timeout: 258 seconds]
drakonis has quit [Quit: WeeChat 2.5]
<eyJhb> Gilfoyle-:imposter!
eyJhb is now known as eyJhb
vika_nezrimaya has quit [Ping timeout: 258 seconds]
__monty__ has joined #nixos-chat
<joepie91> andi-: I feel like ultimately the problem here is that the Nix performance issues are seen as inherent to the model, rather than "not all the optimization work has been done yet"
<joepie91> (and they're probably inherent to /some/ non-zero degree, but not to the degree that we have now)
<andi-> joepie91: yeah, I actually like how nix does it right now. It tries very hard to not have any invalid references in the store. Even if the store path would still not be valid(!?) if things go south. I think yorick also did some work on the downloader recently. So it might use file caches for large files now. Another step would be to download all the things in parallel (up to a specific level) and only
<andi-> "unpack" it once all it's prerequisites are there.
<joepie91> andi-: I feel like there's also some big gains to be had in the parser / import scheduler / whatever
<joepie91> initial parse time is far, far higher than I would expect, in nixpkgs, especially on a HDD
<andi-> The parser needs a makeover if you ask me :D
<andi-> (but thats just because it leaks memory not because I know how to make it any prettiers/faster)
ixxie has joined #nixos-chat
<yorick> andi-: the thing never got merged
<andi-> yorick: not yet ;-)
<yorick> andi-: I don't have high hopes. Serializing nars across a local socket is a neat abstraction but it's not really any sort of optimal
<yorick> you could zero-copy things into the nix store if you were smart about it and ran btrfs or bcachefs
<yorick> (or zfs and you fixed reflinking there)
<andi-> We could do many things. I am not entirely sure what must happend for someone to implement that and to get it merged.
<yorick> well, my patch was needed to get >2GB files to work at all for someone
<andi-> I think it would be a nice intermediate solution to fix things on "low" memory devices
<andi-> without having to provision swap
<yorick> ideally it shouldn't really buffer things in memory in most of the cases where it does
<andi-> I know. it should all follow the Sink/Source thing
<andi-> Or even better with zero copy and the like
<yorick> except sometimes it's a StringSink
vesper11 has joined #nixos-chat
vesper has joined #nixos-chat
vesper11 has quit [Ping timeout: 245 seconds]
<pie_> holy crap github has jump to definition for some things now
endformationage has joined #nixos-chat
infinisi1 has joined #nixos-chat
infinisil has quit [Quit: Configuring ZNC, sorry for the joins/quits!]
infinisi1 has quit [Quit: Configuring ZNC, sorry for the joins/quits!]
infinisil has joined #nixos-chat
drakonis has joined #nixos-chat
infinisil has quit [Quit: Configuring ZNC, sorry for the joins/quits!]
infinisil has joined #nixos-chat
<emily> pie_: haskell-powered, no less
<emily> pie_: thanks for the pointer to expresso, has a lot of things I've wanted in a nix replacement.
AluisioASG has joined #nixos-chat
lopsided98 has quit [Remote host closed the connection]
lopsided98 has joined #nixos-chat
<pie_> emily: yeah something something `semantic`
<yorick> emily: but not on haskell
<yorick> cvuss
<yorick> stuff*
<emily> mhm
<emily> makes sense
<emily> that part is driven by ~business needs~
<emily> just because it makes sense for them to have this one repo written in Haskell doesn't mean they care about the relatively small number of Haskell modules overall :p
<emily> (though I'm sure it'll happen in time)
<emily> see also: facebook's haskell code to deal with PHP
<yorick> pie_: is that like dhall
<pie_> i have nooo ideaaa
<pie_> well to be fair the overall productivity gained by being able to click around in python code might be worth it
<pie_> alternatively its horrible because now people will resort to slow web uis? :P
<emily> nothing's horrible about it, i like it
<emily> the web is a fine platform for annotated interlinked source code
<emily> that's literally hypertext
<emily> running it on github's computers rather than your own is one thing, admittedly, but UI-wise there's nothing wrong with web source browsers imo
<pie_> "i have nooo idea" @ yorick
<pie_> emily: i dont think i was implying there was anyhting inherently wrong with it, besides having to wait for page loads, and even there i was more intending to poke fun at it
<ashkitten> why is there no "cast to network pulse server" thing for android tbh?
<__monty__> tilpner: zfs isn't any more ram-hungry than other filesystems?
<tilpner> __monty__: There's a lot of myths around that, and I don't have definite answers
<tilpner> It undoubtedly profits from large amounts of memory
<tilpner> But so does ext4
<__monty__> Is btrfs particulary more parsimonious with memory use?
<tilpner> I've never used btrfs
ixxie has quit [Ping timeout: 272 seconds]
<tilpner> __monty__: Except if you enable deduplication, then zfs eats ridiculous amounts of memory. Normal usage is fine though
drakonis has quit [Ping timeout: 252 seconds]
<adisbladis> zfs ootb uses a lot of memory
<adisbladis> But it's tunable
<adisbladis> Also the memory looks a bit different to user space
<__monty__> What does that mean?
<__monty__> Zfs is not a fuse fs, is it?
<adisbladis> __monty__: No, but it tend to look like actually used in top/htop
<adisbladis> Which gives the impression that it uses a lot of ram
<adisbladis> But mostly it's just cache
<__monty__> Doesn't that hold for everything? It's like how people complain about browser memory usage.
drakonis has joined #nixos-chat
<adisbladis> __monty__: Well, yeah... I guess it's just even easier to complain about file systems
<adisbladis> Especially when btrfs memory usage looks like cache usage in `free -m` while zfs does not
<__monty__> TIL
<__monty__> Is that because it's in-tree?
<__monty__> I'm not clear on why zfs couldn't make use of such reporting.
<adisbladis> __monty__: I'm not sure either, it's just my observation
<adisbladis> I figured it didn't matter
<adisbladis> It may also have been fixed and my information is old
<drakonis> __monty__: no its because zfs has the spl
<__monty__> spl?
<drakonis> solaris porting layer
<drakonis> it makes zfs alien to any OS
<drakonis> because zfs still makes assumptions inherent to solaris
<__monty__> Oh, ok. TIL again : )
<__monty__> nn, peoples
__monty__ has quit [Quit: leaving]
<adisbladis> __monty__: o/
<pie_> qyliss: experience note on your config: im having a bit of trouble getting heavily modded version of it to work right :P
<pie_> by no fault of yours
<pie_> I was trying to figure out how do something with overlays yesterday but i cant remember right now
<qyliss> pie_: feel free to write me an email
vika_nezrimaya has joined #nixos-chat
<pie_> joepie91: wish nix could be more streamlined
<pie_> the build downtime is real
<pie_> im just complaining because im yak shaving again
<pie_> sigh