<pie_>
also what do you guys think about encouraging people to add documentation files to indivudual packages, for example about build quirks and whatnot? stuff that doesnt necessarily need to be in the main manual
psyanticy has quit [Quit: Connection closed for inactivity]
<thoughtpolice>
gchristensen: I have cached 404s working :)
<samueldr>
yes, I meant to check the resulting build
drakonis1 has quit [Quit: WeeChat 2.4]
drakonis_ has joined #nixos-dev
__monty__ has quit [Quit: leaving]
justan0theruser is now known as justanotheruser
drakonis has quit [Ping timeout: 276 seconds]
<qyliss>
gchristensen: I think soft-deprecating would be fine
<qyliss>
We shouldn't introduce fixed outputs into nixpkgs without careful consideration of whether the software in the closure could ever change the output but still be considered correct, as was the case with cargoSha256
<gchristensen>
yes!
aanderse has joined #nixos-dev
<gchristensen>
having a "oh this adds a new f-o fetcher! this needs extra review!" is good
* samueldr
thinks about the fonts with FOD
<gchristensen>
but I think eelco's plan was actually to remove them from the language :)
<pie_>
eelco seems to like moving things into the evaluator?
<pie_>
*interpreter
<gchristensen>
what makes you say that?
<pie_>
It just kind of feels like that from random browsing I've done, but I'm not sure it's true.
<gchristensen>
right
<gchristensen>
well if you want to guarantee an expression isn't sneaking around purity rules, making it impossible is a good way to do it
<pie_>
sure
<gchristensen>
is more where he's coming from
<pie_>
but I would prefer the language support impurity in a well controlled flexible manner as opposed to no user expandability
<gchristensen>
okay
<pie_>
ok to be clear im saying i want to be able to do controlled impure stuff, but maybe im in the wrong here
<gchristensen>
you probably don't want Nix to be involved in doing impure stuff. graph traversals get ugly ugly
<pie_>
so do we need another thing on top of nix? :P
<pie_>
s/:P//
<qyliss>
What sort of use-cases are you thinking of?
<pie_>
not sure :(
<qyliss>
in other words, what is it that makes you wish you could do impure stuff?
<pie_>
I kind of want to just write scripts for everything as opposed to running stuff in a shell
<pie_>
maybe thats a bad workflow
<pie_>
eh, nevermind, I only have vague inklings here that would be high effort on your parts
<pie_>
well I did have that one use case I mentioned earlier, but maybe you can explain what is wrong with it
<pie_>
https://github.com/kamilchm/go2nix generates a lock file for a go project and then uses buildGoProject or whatever it was for building it, and I wrote a fixed output derivation that ran go2nix for me and gave me the .nix lock files, which I then used via IFD
<pie_>
(lots of hash changing ensues)
<qyliss>
Why wouldn't you do a shell script for that?
<qyliss>
Ultimately Nix is probably generating a shell script anyway
<pie_>
Hm. Maybe it's just preferring Nix as my top level and not a shell script. (?)
<pie_>
otherwise I end up reimplementing the nix sandbox
<pie_>
ok maybe i just want the sandbox
<samueldr>
are you using the nix sandbox to run end-user software?
<pie_>
sure why not
<pie_>
i run batch jobs with it sometimes that dont need x11
<pie_>
blame clever
<pie_>
:P
<pie_>
*that dont need user interaction (forgot about xvfb for a second there)
<samueldr>
ah, yeah, that's fine, data goes in, data goes out
<samueldr>
I was thinking more about like... your irc client or some crazy use case
<pie_>
yeah i didnt really answer in the spirit of the question
<pie_>
hmm i havent tried that yet :P
<samueldr>
it wasn't a suggestion :^)
<pie_>
"im gonna need more nixbld users"
<pie_>
microkernels are for schmucks, need more bind mounts
<pie_>
*namespaces (/ i dont even)
<pie_>
samueldr: I tried forwarding sockets over cntr but it didnt work