endformationage has quit [Ping timeout: 248 seconds]
aminechikhaoui has quit [Quit: Ping timeout (120 seconds)]
aminechikhaoui has joined #nixos-chat
<eyJhb>
Is it just me, or is Github support really slow?
jasongrossman has quit [Ping timeout: 250 seconds]
pie_ has quit [Remote host closed the connection]
pie_ has joined #nixos-chat
pie_ has quit [Ping timeout: 248 seconds]
<qyliss^work>
eyJhb: they're usually really fast for me
<qyliss^work>
gchristensen: we can't recruit anyone else to join me and my manager in my job; been trying for months
<qyliss^work>
surprises me, because from my point of view it's, like, the perfect job
<gchristensen>
qyliss^work: right?
<gchristensen>
manveru: yeah? I can put you in touch. ... :)
<eyJhb>
qyliss^work: It has currently taken around two weeks, and I have gotten a single reply 1 week ago... And waiting for another
<qyliss^work>
that's extremely unusual from my PoV. Usually takes a few hours for me at most.
<qyliss^work>
Maybe more if it's something complicated they have to look into
pie_ has joined #nixos-chat
<eyJhb>
Basically looking to get a unlimited team subscription for a org I have.. Which is opensource it MOST aspects.. And I cannot pay for a team plan, because.. It wouldn't make sense
<eyJhb>
Basically, there is no plan for what I need :(
<eyJhb>
Unless I get this approved
<gchristensen>
ah, are you a registered nonprofit anywhere?
<eyJhb>
Nope, currently not. And I don't know if it would make sense, because it is a subproject in my company. The real trouble is, that it is part of a security platform, where the challenges for the platform will be at Github. But the newest challenges would need to be private, to not leak the answers (but would be public, after used). So the trouble is, there would be a lot of contributors in private repos,
<eyJhb>
and some might be for 1 day, others for a little more...
<eyJhb>
So yeah, it is kind of a weird edge case. If I could, I would just make everything public. But that doesn't work well, for active CTFs...
<gchristensen>
maybe gitlab would be a better choice
<eyJhb>
Considering it, but I like the way Github functions.. But if the decline it, I might have to use Gitlab instead.... :/
<gchristensen>
yeah, I think "a subproject inside a company" is going to be a tough sell, even if the use case is novel
<qyliss^work>
do they pro-rate org members?
<qyliss^work>
can you add people for a single day and pay for that one day?
<eyJhb>
gchristensen: yeah... But hoping I can pull a little on the academic part of it, since I am at Aalborg University, and it would be the main place..
<eyJhb>
I don't think so, it is pr. user pr. month
<qyliss^work>
Lots of things that are /user/month will actually pro-rate
<qyliss^work>
I know CodeClimate does, because I get the emails about our work's huge subscription every month
<eyJhb>
Might be a option. It just generally wouldn't be that nice, to have to pay for it. The only thing I need is enough money, so that I can quit my student job and still survive...
<eyJhb>
Maybe they would take the same deal, as the private individual repos? 3 collaborators on private
arianvp has quit [Quit: WeeChat 2.4]
Jackneill has quit [Ping timeout: 246 seconds]
Jackneill has joined #nixos-chat
<eyJhb>
Have anybody played with Docker and where it saves environment variables
<eyJhb>
?
<gchristensen>
sure
<eyJhb>
When I spin up a container, is there then a safe way, that I can destroy the environment variable at boot? Using `./init.sh` does not work
<eyJhb>
E.g. I start a instance - `docker run --rm -it -e "FLAG=test" alpine`, i `unset FLAG`, but to it again, FLAG will of course still be there
<eyJhb>
So nay place, that I can remove it? Or maybe override it?
<gchristensen>
I'm not really understanding, and the parts I was familiar with is the format of the container image itself
<gchristensen>
I don't really know how to use Docker anymore... just how to build their file formats.
<eyJhb>
Ahh... Yeah, this goes somewhat beyound that. Currently I have a dockerfile with a `ENV FLAG something`, and I need to be able to unset it again (globally) on the imag when it starts
<andi->
You could probably derive an image that unsets it?
<eyJhb>
andi-: I need to use it once the image starts, but then nothing more
<eyJhb>
That would only work for the current session, if I attach to it or spawn a new shell it would be back
<eyJhb>
Or does it.. Might need more testing. I just thought it would save it some place
<gchristensen>
sounds like a better thing to do is make a new image without it set ... :P
<eyJhb>
Yeeaahh.. But that would require recompiling everything, to change the flag :p
<andi->
ha, the hours I wasted waiting for docker rebuilds because I flapped an unrelated bit :D
<andi->
*flipped
<eyJhb>
... Having a OpenCV container, I can confirm....
<gchristensen>
graphs *clap* not *clap* layers
<eyJhb>
But at least I get to waste uni time
<gchristensen>
you could use Nix to build your docker containers, and then you'll spend much less time waiting for docker's layers to be happy
<andi->
use the amazing layered docker build from gchristensen ? :)
<eyJhb>
Would be stupid to force people to use it :p
<gchristensen>
it would?
<eyJhb>
Open source project where I would like people to contribute with challenges? Getting them to use Docker in itself is a challenge
<eyJhb>
And there are Windows people too.................
<andi->
so we need a Dockerfile parser in Nix so that people with Nix are faster and have the better(tm) builds and the rest of the world waits and drinks coffee?
<eyJhb>
^^^^
<andi->
(says me while waiting for GitLab CI of Debian builds..)
<andi->
(and drinking coffee)
<eyJhb>
I am basically just watching my group burn, trying to get some Golang and webcam to work
<eyJhb>
I just heard them say nmap, wtf is happening
<andi->
did they say JTAG or DTRACE yet?
<eyJhb>
No no, not yet!
<eyJhb>
I found the legacy way it worked...
pie_ has quit [Remote host closed the connection]
<eyJhb>
Just got the best argument for Windows > Linux. I can use Wifi on Windows... What
<Taneb>
That used to be a lot more of a thing a decade ago
<Taneb>
...it was the main reason I didn't stick with Haiku back when I was a hipster nerd teenager
waleee has joined #nixos-chat
endformationage has joined #nixos-chat
pie_ has joined #nixos-chat
<pie_>
its kind of annoying that writeshellscriptbin puts things in a bin subdirectory, i think id much more often would have wanted a top level sh file
<pie_>
but maybe thats just me
<srhb>
pie__: Um. That's sort of the point of the bin, isn't it?
<tilpner>
You can use writeScript with runtimeShell
<pie_>
oh huh we have a writeshellscriptbin but no writeshellscript
<pie_>
i mean i can see the source code of writeshellscriptbin, its pretty simple
<pie_>
tilpner, i guess youre right
<pie_>
though i dont see why we couldnt have a writeShellScript, we seem to like this style of lib functions anyway :p
<tilpner>
We could
<pie_>
im not quite sure what the nice way to add it next to writeShellScriptBin would be without duplicating code
<pie_>
besides making a helper function - i dont want to add another parameter
<pie_>
looks like someone forgot to update the docstring when copy pasting code?
<pie_>
basically i flailed around a bunch trying how to do a python.withPackages style thing, and i still dont have all the architectual ideas down, but then i found lib.makeExtensible and it seems to have covered what i wanted so far
<pie_>
the overall approach:
<pie_>
i compose several overlays,
<pie_>
i call 0_base with pkgs in scope and it adds it to .pkgs, and to its own callPackage's auto scope, so the overlay is otherwise empty and provides general package set infra
<pie_>
i put some misc stuff in 1_config and 2_util, its kind of ad-hoc but it seems to work ok
<pie_>
and i put the enumeration of packages in 3_packages.nix and then the actual package definitions in 4_packages/
<pie_>
(for package sets with larger amounts of packages ive been considering experimenting with automatic enumeration of packages but that might have some drawbacks and i havent worked on it other than writing a really simple implementation)
sir_guy_carleton has joined #nixos-chat
<pie_>
(if you have autoenumeration though you can use the same package set infrastructure for everything tho (?))
<pie_>
(i mean for 3_packages)
<pie_>
and this setup lets you do (rootPackage.extend anotherOverlay).withPackages (p: with p; [ myAnotherPackage ]), which is what i wanted
<pie_>
hmm i think this repo doesnt have all the stuff i did
<pie_>
but i dont have the other repo up yet
<pie_>
i just do something like let packages = import packages; in { agda = packages.agda } // packageSet for convenience
<pie_>
so then you get the nice agda.withPackages thing.
<Taneb>
I'm not sure I fully understand
<pie_>
im not sure it fully makes sense :P
<pie_>
ah sorry, more like: let packages = import packages; in { agda = packages.agda } // packages
<Taneb>
And it's definitely a different conceptual layer to what I was planning
<Taneb>
Not completely but at least largely orthogonal
<pie_>
and packages is the composition of the layers from earlier, the packages layer containing a .agda
<pie_>
do you think the approach is any good though? i havent asked for much input yet
<Taneb>
It feels like it has potential
<pie_>
most of the code usually ends up in 1_util
<pie_>
and lib/
<pie_>
possible drawbacks: also infinite recursion problems are not fun to debug
<Taneb>
That they are not
<pie_>
but i think you get that with and of this fix based stuff
<pie_>
*any of
drakonis has joined #nixos-chat
drakonis_ has quit [Ping timeout: 245 seconds]
<pie_>
ok i might be misunderstanding something
<pie_>
can i not just make a /nix/store/hash-something.sh instead of a directory?
<gchristensen>
you can
<pie_>
no wait thats definitely possible
* pie_
pokes around some more
<pie_>
gchristensen, thx
<gchristensen>
I prefer that whenever possible because "${script}" is certain to be correct in that case, where "${script}/bin/whatever" is not
<pie_>
ugh writetextfile is doing something weird
<pie_>
mv: cannot move '/build/.attr-0' to '': No such file or directory
<colemickens>
and lid handling? anything special there? do you manually disalbe the internal display?
<gchristensen>
so ... that seems to "just work"?
<colemickens>
aha, interesting. I think it used to for me.
<colemickens>
And there's another specific output-handling bug I'm hitting. I wonder if there's a weird interaction between them.
<colemickens>
Time to dig in. Thanks again, if I'm good I'll try to put some of this in a wiki page later. There are dozens of us! (well, a few at least)
<gchristensen>
:P
<gchristensen>
on one hand, I'm surprised at how easy this has been to set up. on the other hand, also not exactly
<colemickens>
when you say that, you mean "drag over with mouse, ctrl+Shift+C" style works? that's always been there for me in termite at least. Though I'm not a huge fan of termite and its weird TERM + ssh.
<colemickens>
I'm actually using alacritty in Windows with the new TTY system pretty effectively. I'll have to give it a shot in Linux again.
<gchristensen>
right
<gchristensen>
(though, alacritty doesn't do box selection like sakura does, yet)
<gchristensen>
(hrm, I should have deleted these 75,000 messages in the "grahamcofborg" mail folder before syncing...)
<colemickens>
Workspace and extensions run remote. Local UI becomes a thin client. Containerized dev/build environment (imagine with lorri... :D)
<gchristensen>
:o
<colemickens>
Or if Theia can combine forces and become a supported web UI, since they've sort of been building the same model, with the same base codebase.
<infinisil>
TIL that you can nagivate through github repo's folder structure with vim key bindings
<eyJhb>
If I were to make 200 requests every 5 secounds to a SQLite database which needed to perform 5 deletes and 5 inserts, or something like that. Then.. How bad would it die?
<simpson>
eyJhb: How many writes to disk? (40/s?) How many writes can your disk support?
<eyJhb>
I am very much in doubt, on what DigitalOceans specs are. But a cheap droplet from them would run it
<sphalerite>
from a completely gut-feel estimate, I'd think that that would be perfectly manageable on lower-middle-end consumer hardware nowadays.
<eyJhb>
Maybe a db in memory, which would be saved to disk every x secounds? It is important to note, the database would not increase in size, but stay somewhat the same (hence, delete and insert)
<simpson>
Ah, sure. They have SSDs, which can be fast.
<eyJhb>
I reality, more like "update, and if you do not exist in this request I just got, you get deleted, and if you are new, you will be added"
<eyJhb>
Great!
__monty_1 has joined #nixos-chat
disasm| has joined #nixos-chat
__monty__ has quit [Ping timeout: 248 seconds]
disasm_ has quit [Ping timeout: 248 seconds]
hedning_ has quit [Ping timeout: 248 seconds]
Kim has quit [Ping timeout: 248 seconds]
__monty_1 is now known as __monty__
endformationage has quit [Ping timeout: 245 seconds]
Kim has joined #nixos-chat
<MichaelRaskin>
gchristensen: to be absolutely fair, when I had a NixOS installation, it wouldn't help me any to find out whether something is included in the default installation.
<gchristensen>
yeah, that wasn't really what they were asking
<gchristensen>
but I hear that
<gchristensen>
also, you read a lot of my tweets for someone who doesn't follow me
<joepie91>
eyJhb: main thing to keep in mind is that SQLite is bad at write concurrency
<gchristensen>
joepie91: ...tell that to expensify...
<MichaelRaskin>
gchristensen: I said that I read your tweets quite often!
<MichaelRaskin>
Not following is more about not having a Twitter account
<gchristensen>
ah I see :)
<gchristensen>
I imagine you can appreciate the vampiric behaviors I was referring to, though
<__monty__>
So people who read what you say owe you a follow?
<samueldr>
infinisil: do you know about the `t` key shortcut in a github repo? (also works with gitlab)
<samueldr>
when in the files listing
<infinisil>
Ah yeah I've accidentally pressed it occasionally
<MichaelRaskin>
I am kind of scared by the enthusiasm of «let's just add string interpolation to unquoted URLs» to be honest.
<gchristensen>
wait what
<pie__>
wait what
<MichaelRaskin>
Oh, I got some comments on my RFC PR
<MichaelRaskin>
I would say that after seeing pushback on some mere builtin proposals, I wouldn't expect the same person to propose such a syntax extension
<MichaelRaskin>
But what do I know
<MichaelRaskin>
Want a link to the comment?
<gchristensen>
yea
<samueldr>
that is understandable, if the premise is "urls aren't acting like strings enough, even though they're typed as strings at runtime"