gchristensen changed the topic of #nixos-chat to: NixOS but much less topical || https://logs.nix.samueldr.com/nixos-chat
<lovesegfault> gchristensen: so, with the systemd-run thing, how do you get the processes reaped when the WM exists?
<lovesegfault> s/exists/exits/
<lovesegfault> I exited out of sway only to see all my windows were still running b/c they're not child procs of it anymore
<infinisil> lovesegfault: I think it spawns everything in a "slice"
<infinisil> So if you know which slice it is, you can systemctl stop that
<infinisil> I think
<lovesegfault> infinisil: app.slice it seems
<samueldr> can a wayland app survive the compositor exiting and a new one being launched?
<samueldr> if so, that's a net advantage to launching outside the PID of the composito
<lovesegfault> samueldr: having each proc in its own cgroup
<samueldr> yeah
<samueldr> that's not the question
<lovesegfault> so I can limit slack's memory usage
<samueldr> I know what you're doing
<lovesegfault> and AFAIK the answer to the first question is "no"
<lovesegfault> like, the proc will survive, but when you restart the compositor there's no way to recover the app
<samueldr> uh... so no way to switch to another "wm" (read: compositor) like you can on X11, :(
<lovesegfault> AFAIK
<samueldr> also, bad resilliency on crash
<lovesegfault> I'm writing a PR to get command-not-found to suggest nix-shell instead of nix-env. what word should I use instead of install?
<samueldr> what's the sentence?
<lovesegfault> The program ‘$program’ is currently not installed. It is provided by
<lovesegfault> several packages. You can install it by typing one of the following:
<lovesegfault> and slight variations of that
<samueldr> hm
<lovesegfault> The program ‘$program’ is currently not installed. You can install it by typing:
<lovesegfault> nix-env -iA nixos.$package
<lovesegfault> etc
supersandro2000 has quit [Disconnected by services]
<lovesegfault> I'm thinking this:
supersandro2000 has joined #nixos-chat
<samueldr> you can [get] a shell with it by typing:
<samueldr> but get feels bad
<lovesegfault> The program '$program' is currently not available. You can make it available in a shell by typing:
<samueldr> it'd be nice to also say it's ephemeral
<samueldr> or temporary
<lovesegfault> maybe `is not in your environment` instead of `is currently not available`?
<samueldr> good idea
<lovesegfault> sure, "You can make it available in an ephemeral shell by typing:"
<samueldr> it might be available already in their nix store :)
* lovesegfault nods
<lovesegfault> okay, trying that
<lovesegfault> TIL if you add NIX_AUTO_INSTALL in your path when you run a command you don't have it automatically installs with nix-env
<lovesegfault> terrifyintg
<lovesegfault> and if you set `NIX_AUTO_RUN` it automatically runs it in a shell for you
<lovesegfault> that one is neat
<samueldr> whew, looks like it's an ancient feature :)
<cole-h> lovesegfault: "The program '$program' was not found. You can use it by running the following:" ?
<lovesegfault> not found may makes it sound like it's not in nixpkgs
<lovesegfault> s/makes/make
<cole-h> Why would it make it sound like that?
<cole-h> Nixpkgs isn't really involved at that point, so I don't think it would make sense to think that way in that context
<lovesegfault> Idk, I'm trying to think from the POV of someone's first interaction with Nix
<lovesegfault> maybe you're right
<lovesegfault> but I think "not found" leaves out "not found" where?
<cole-h> "not found in PATH"
<lovesegfault> in my env? in my nix store? in nixpkgs? in my available caches?
<lovesegfault> right, so I'm trying to put that in the sentence to make it extra clear
<lovesegfault> Maybe "The program '$program' is not in your PATH. ...`
<cole-h> Sure
rajivr has joined #nixos-chat
<{^_^}> #108184 (by lovesegfault, 16 seconds ago, open): nixos/command-not-found: don't suggest nix-env
<cole-h> I really wish we could stop using "fancy" quotes...
<abathur> ?
<cole-h> ‘$program’
<cole-h> Those quotes
<cole-h> Hard to grep for
<abathur> oh, huh; is that a perl thing, generally? if so I guess I haven't been reading perl very carefully :p
<cole-h> Don't think so
<cole-h> Pretty sure it's just "fancy" stuff that was added once upon a time and never removed
<abathur> ah, hmm
<samueldr> not a perl thing at all, just the convention initially used
<supersandro2000> cole-h: sometimes you can escape them with "'"
<supersandro2000> " ' "
<infinisil> I think it may be a localization and smart-quotes thing
<samueldr> it's about using the right symbols
<samueldr> " and ' are poor approximations
<cole-h> If I want to `sed` something between smart quotes, I can't just do `sed "s/'.*'/'lolololol'/"` -- I have to do `sed "s/(find smart quote open and paste).*(find smart quote close and paste)/(find smart quote open and paste)lolololol(find smart quote close and paste)/"`
<infinisil> Some keyboards/editors might insert these localized marks automatically
<samueldr> cole-h: compose <" and compose >", <' and >'
<samueldr> for the french guillements, « << and » >>
<cole-h> And if I don't have a compose key?
<cole-h> :P
<samueldr> get that fixed asap?
<samueldr> how are you even typing things?
<cole-h> samueldr: With my fingers, duh
<cole-h> and my thumbs
<samueldr> I walked right into that, didn't I?
<cole-h> :D
<supersandro2000> infinisil: the only program I know that does that is word
<supersandro2000> and most wordpress websites for which I hate them
<infinisil> Iirc macOS' default text editor does so too
<supersandro2000> default keyboard is dumb
<abathur> also neat https://www.gnu.org/prep/standards/html_node/Quote-Characters.html#Quote-Characters (in the sense that I've oft wondered why, in older documents...)
ky0ko has joined #nixos-chat
<MichaelRaskin> As a TeX user I missed the fact that there is anything to wonder about in the case of ``'' quoting…
<supersandro2000> MichaelRaskin: https://github.com/Mic92/tex2nix
<supersandro2000> JFYI
<lovesegfault> I love that TeX's answer to package management was "just have all the packages in the world on your system always"
<lovesegfault> QED
<MichaelRaskin> Well, I already have a list of what TeXLive packages I need for various articles I actually need to edit and recompile (as an expression, of course)
<MichaelRaskin> lovesegfault: CTAN doesn't have that approach, though.
<lovesegfault> MichaelRaskin: right, but everyone just uses texlive IME
<MichaelRaskin> Really? Not Debian, which splits it in large chunks, but still splits it quite a lot? Not MikTeX with its GUI for picking CTAN pieces?
<lovesegfault> MichaelRaskin: I mean individuals, not distros. Whenever you're being taught how to use TeX it's just "install all of texlive"
<lovesegfault> even if you run a distro that splits things up, like debian or Arch, you'll still find people commonly recommending you just "install everything"
<lovesegfault> Again, IME
<MichaelRaskin> It's actually a nontrivial amount of work to install everything of TeXLive in Debian
<MichaelRaskin> I think there are no metapackages for _that_
<MichaelRaskin> It's another thing that if a person even needs to be told, there is an incentive to get them to install everything to increase mean time to next question
<lovesegfault> :D
<MichaelRaskin> There are enough smooth-curve tutorials that if you are being personally taught…
supersandro2000 has quit [Ping timeout: 246 seconds]
supersandro2000 has joined #nixos-chat
tilpner_ has joined #nixos-chat
supersandro2000 has quit [Quit: The Lounge - https://thelounge.chat]
tilpner has quit [Ping timeout: 240 seconds]
tilpner_ is now known as tilpner
supersandro2000 has joined #nixos-chat
Emantor has quit [Quit: ZNC - http://znc.in]
Emantor has joined #nixos-chat
<lovesegfault> supersandro2000: you around/
<lovesegfault> *?
<supersandro2000> whats up`
<supersandro2000> =
<supersandro2000> ?
<supersandro2000> lovesegfault: ping
ajs124 has quit [Quit: killed]
das_j has quit [Quit: killed]
das_j has joined #nixos-chat
ajs124 has joined #nixos-chat
<lovesegfault> supersandro2000: I ended up merging it myself
<lovesegfault> since I realized I had an old review and nothing had changed
<lovesegfault> but was going to ask you for a review
<supersandro2000> no problem
<supersandro2000> I just found THE best thing since sliced bread
<supersandro2000> all options, context aware 107KB!
<supersandro2000> wait... Why am I cloning this... I could just add it to nixpkgs
<supersandro2000> oh well to lazy right now... already way to much of the merge train
<supersandro2000> and before anyone asks: that won't really slow down the shell. that only happens if programs wrap their static and never changing completion script behind npm
srk has quit [Ping timeout: 240 seconds]
srk has joined #nixos-chat
eyJhb has quit [Quit: Clever message]
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
eyJhb has quit [Changing host]
eyJhb has quit [Remote host closed the connection]
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
eyJhb has quit [Changing host]
<infinisil> TIL that the topmost layer of our skin is essentially dead: https://biology.stackexchange.com/a/97350
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
eyJhb has quit [Changing host]
<abathur> <3 infinisil
<{^_^}> infinisil's karma got increased to 396
supersandro2000 has quit [Read error: Connection reset by peer]
supersandro2000 has joined #nixos-chat
<supersandro2000> did I just fork bomb myself by export GITSTATUS_DAEMON_PID=
<supersandro2000> -_-
ky0ko has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
<samueldr> infinisil: a whole new spin on "you're dead to me"
ky0ko has joined #nixos-chat
<lovesegfault> somehow every time I try to overlay pythonPackages I end up in a hole
<supersandro2000> lovesegfault: https://nixos.wiki/wiki/Overlays and then at the bottom python overlays
<lovesegfault> supersandro2000: I know, I'm doing that
<lovesegfault> but I still get a failed build saying it can't find confuse
endformationage has quit [Quit: WeeChat 2.9]
<supersandro2000> try super.callPackage isntead of pyself
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
eyJhb has quit [Changing host]
<lovesegfault> supersandro2000: no difference
<supersandro2000> you sure it uses the overlay?
<lovesegfault> yes, b/c I see build of beets happening
<lovesegfault> it's curious
<hexa-> do we have another beets issue? :O
<lovesegfault> even if I replace `self.python3Packages` with `python3.pkgs` by making things `rec` it still doesn't work
<lovesegfault> hexa-: no, I'm working on getting the HEAD packaged
<lovesegfault> since the releases are so slow
<lovesegfault> I want to have `beets-unstable` in nixpkgs
<hexa-> have you complained to them already? :P
<lovesegfault> Yes, Adrian is just a busy guy
<hexa-> sounds like something for your NUR repo honestly
<lovesegfault> He'll cut a release when he has time, no point in waiting
<hexa-> are you missing any particular feature?
<lovesegfault> yes, I want a new plugin
<hexa-> like I just do `beet import`, have it fix tags, grab coverart and update mpd
<hexa-> do tell
<lovesegfault> my beets library is... extensive
<hexa-> error: Package ‘perl5.32.0-Glib-Object-Introspection-0.048’ in /etc/nixos/nixpkgs/lib/attrsets.nix:344 is marked as broken, refusing to evaluate.
<hexa-> *sigh*
<hexa-> pretty sure I didn't request this package directly
<lovesegfault> so I need the `unimported` plugin to check, from time to time, if some nontracked file made it's way into my lib
<lovesegfault> and that is only on head
<hexa-> lovesegfault: I only have like 150G of music :)
<hexa-> oh ok
<lovesegfault> Tracks: 15383
<lovesegfault> Total time: 6.6 weeks
<lovesegfault> Approximate total size: 565.6 GiB
<lovesegfault> Artists: 1434
<lovesegfault> Albums: 1127
<lovesegfault> Album artists: 610
<lovesegfault> :P
<hexa-> Tracks: 6507
<hexa-> Total time: 2.2 weeks
<hexa-> Approximate total size: 131.9 GiB
<lovesegfault> I'm also about to bite the bullet and do something I've been waiting to do forever
<lovesegfault> re-import all my music using discogs to tag instead of musicbrainz
supersandro2000 has quit [Read error: Connection reset by peer]
<hexa-> that will take days :p
<lovesegfault> I know, I'm alright with that
<hexa-> I wonder if you'll run into rate limiting
<lovesegfault> perhaps
supersandro2000 has joined #nixos-chat
eyJhb has quit [Quit: Clever message]
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
eyJhb has quit [Changing host]
eyJhb has quit [Remote host closed the connection]
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
eyJhb has quit [Changing host]
cole-h has quit [Ping timeout: 240 seconds]
lunc has quit [Ping timeout: 240 seconds]
eyJhb has joined #nixos-chat
waleee-cl has quit [Quit: Connection closed for inactivity]
eyJhb has quit [Quit: Clever message]
eyJhb has joined #nixos-chat
eyJhb has quit [Excess Flood]
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
<lovesegfault> I need a source file available on this website written in PHP
eyJhb has quit [Remote host closed the connection]
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
eyJhb has quit [Changing host]
neeasade has joined #nixos-chat
<sphalerite> lovesegfault: hm, why that? What's the advantage of discogs?
<lovesegfault> sphalerite: I have a lot of obscure albums that aren't on musicbrainz
<lovesegfault> but are on discogs
<lovesegfault> and migrating them can be a HUGE pita
<lovesegfault> I just don't have the patience
<sphalerite> ah, fair
neeasade has quit []
lunc has joined #nixos-chat
<patagonicus> Heh. While I knew it before, it finally clicked that <nixpkgs> is just as much a path as ./ and I can do stuff like `lib = let foo = <nixpkgs> in import (foo + "/lib");`
<eyJhb> patagonicus: Yeah it is quite nice! But one can quickly screw it up...
<patagonicus> Oh, for sure. My use case is pinning nixpkgs in CI, but giving the option of using nixpkgs from NIX_PATHS locally if desired (and I can no toggle that via --arg pinned false).
<patagonicus> Next step: getting caching of the nix store to work in GitLab CI. That's going to be interesting.
__monty__ has joined #nixos-chat
cosimone has joined #nixos-chat
cosimone has quit [Quit: cosimone]
cosimone has joined #nixos-chat
thibm has joined #nixos-chat
waleee-cl has joined #nixos-chat
cole-h has joined #nixos-chat
spudly1 has quit [Ping timeout: 265 seconds]
cole-h has quit [Ping timeout: 256 seconds]
<infinisil> lovesegfault: hexa-: 3415 tracks for 2.0 weeks total time for me, only 42.0 GiB :)
<hexa-> infinisil: get more flac :p
<infinisil> Honestly it makes no sense for me to get flac's
<hexa-> yeah, kidding. But that's probably the difference.
* infinisil nods
<infinisil> > 42.0 * GiB / 3415
<{^_^}> 1.32056e+07
<infinisil> > (42.0 * GiB / 3415) / MB
<{^_^}> 1.6507
<infinisil> Average MB per song
<infinisil> > (131.9 * GiB / 6507) / MB # hexa-'s library
<{^_^}> 2.72066
<infinisil> > (565.6 * GiB / 15383) / MB # lovesegfault's library
<{^_^}> 4.9349
<hexa-> that sounds somewhat wrong
<infinisil> Oh hmm yeah
<hexa-> it should be more around 20 MB per song
<infinisil> Yeah..
<hexa-> and 1.65 MB per track also doesn't sound great :)
<infinisil> > GiB
<{^_^}> 1.07374e+09
<infinisil> > 1024 * 1024 * 1024
<{^_^}> 1073741824
<infinisil> That looks correct
<infinisil> > 44.0 * 1024.0 * 1024.0 * 1024.0 / 3415.0
<{^_^}> 1.38344e+07
<infinisil> > 44.0 * 1024.0 * 1024.0 * 1024.0 / 3415.0 / 1000.0 / 1000.0
<{^_^}> 13.8344
<infinisil> Huh
<infinisil> > MB
<{^_^}> 8e+06
<infinisil> Oh
<infinisil> I think I might have made a mistake in these units, confusing bits with bytes
* infinisil goes to #bottest
<infinisil> > (42.0 * GiB / 3415) / MB
<{^_^}> 13.2056
<infinisil> There we go, should be fixed now
<infinisil> > (131.9 * GiB / 6507) / MB # hexa-'s library
<{^_^}> 21.7653
<infinisil> > (565.6 * GiB / 15383) / MB # lovesegfault's library
<{^_^}> 39.4792
<infinisil> That looks better
<hexa-> yep
<patagonicus> Wow. The GitLab CI cache was not working, because GitLab does a git clean -ffdx, which then deletes the cache file. Seems kinda … stupid.
<abathur> huh, volth lives, at least on discourse
ece has quit [Ping timeout: 264 seconds]
<das_j> patagonicus: Did you tell it where the cache is? You have to specify that in your gitlab-ci.yml
<infinisil> Damn this is really cool, looks almost like CGI: https://www.youtube.com/watch?v=fn3KWM1kuAw
<infinisil> Boston Dynamics robots doing a dance together
<patagonicus> das_j: Yeah, I had the cache specified. But I'm just running it locally, no cache server, with the "docker" executor. But until I also set GIT_CLEAN_FLAGS to not delete the cache it would always delete the file before any of my stuff ran.
<patagonicus> I'm really not sure how the cache is supposed to work - maybe you're not supposed to have it in the working directory? But I remember reading that it has to be.
<sphalerite> infinisil: aw yiss cute military tech is the best, almost helps you forget that this stuff is designed to help kill people!! :D :D :D
<infinisil> -.-
<infinisil> Of course everything can be used for bad things as well. But these robots are designed to help warehouse operations
<das_j> patagonicus: We have a node project where the node project is in the working directory, but a directory itself. But we are also using the shell executor
<das_j> *node cache
<patagonicus> Hmm
<__monty__> Tbh, I prefer the pillaged NASA money going into practical robotics over PR or conventional weaponry.
<das_j> patagonicus: We use `cache: { key: "condor-node-modules", paths: [ "node_modules/" ] }`. Maybe that helps
<patagonicus> That's roughly what I was doing, didn't work. Are you sure it's actually using the cache? The thing is that these caches are just for performance, so the jobs should succeed even if the caches not there. :D
<patagonicus> To be fair, my GitLab is on an old version, but the runner is not.
<joepie91> if I got a dollar for every time "should succeed even if the cache is not there" turned out to be "oops, someone decided to use the cache as a critical persistent datastore without telling anybody"...
* joepie91 glares extra hard at all the JS devs stuffing pretend-not-globals into the module cache
<patagonicus> joepie91: At least I don't have that problem, because all I'm caching is nix store paths, so if there not there, it shouldn't matter. :)
<andi-> unless you produce those paths impurely or via some weird fixed output stuff that wasn't really a good idea ;)
<patagonicus> Well, it's working now, although I'm not caching everything. But I am caching the extracted, pinned nixpkgs as well as most build dependencies, so it's not downloading much. I can live with that. Especially because I probably ran the pipeline more often to debug this than I will for the rest of the lifetime of the project. :D
<patagonicus> andi-: Nah, it's just one buildGoModule with vendored deps and a shell for linting/testing.
<andi-> vendored deps as in they are in the source checkout or are vendored using the nix tooling? :)
<patagonicus> Source checkout, as I want this to be easily buildable without nix, just Go tooling.
<patagonicus> Ok, I get the feeling that the cache bit doesn't do anything and the "caching" only works because I told Gitlab not to clean up in between runs.
rajivr has quit [Quit: Connection closed for inactivity]
<hexa-> python-mocket just went from 3.9.4 to 3.9.35 in which they removed a py2 compat layer
<hexa-> I hope it's just a typo
<hexa-> but most probably it reflects minimal supported python version is now 3.5
thibm has quit [Quit: WeeChat 2.6]
<joepie91> wat
<joepie91> that is a terrible way to version things
<srk> patagonicus: is there some guide about running nix builds (and possibly qemu tests) on gitlab pipelines?
<patagonicus> srk: no idea, but outside of the caching thing it's really simple. Basically just image: nixos/nix:2.3.6 and nix-build --no-build-output. Although my default.nix uses a pinner nixpkgs, otherwise you need to nix-channel --add first.
<srk> cool
ece3 has joined #nixos-chat
andi- has quit [Quit: WeeChat 2.8]
andi- has joined #nixos-chat
endformationage has joined #nixos-chat
<sphalerite> infinisil: nah, robots for warehouses don't need fancy leg stuff, they can go on wheels. Boston Dynamics has mostly been funded by DARPA as far as I know.
<sphalerite> unless I've missed the development of cross-country warehouses with rough terrain :p
<__monty__> I think firefighting's one of the applications of the bipeds?
<sphalerite> ah yes, for the famous well-funded fire departments
<__monty__> Well, you need expensive robots first before they can start getting cheap enough for fire departments.
<samueldr> one of the main real-use things for boston dynamics quadrupeds is climbing stairs
<sphalerite> energizer: ooh that's the https://www.youtube.com/watch?v=0Pp7ilNkxhQ right?
<samueldr> so you can have it walk through your oil rig, including stairs, to do automatic inspections
<samueldr> or a building in construction, to update a map daily, or more often, to track progress
<sphalerite> oh yeah oil and gas, the pioneers of doing good for humanity.
<sphalerite> I'm sorry
<sphalerite> I should just shut up and go to bed :)
<sphalerite> energizer: oh wow, I think I missed that xkcd! hahaha
<samueldr> nah, I was mainly thinking about stated use cases
<samueldr> and anything that can reduce (or eliminate) the risks is good, since we're not close to completely eliminitating their uses yet :(
<samueldr> in a realist point of view: better have additional checks than none at all on those bad actors
cosimone has quit [Quit: cosimone]
<joepie91> one major problem that that approach often presents for activists (in general): sometimes the only way to shut down a bad actor is to focus on something relatively minor because the more serious problem is difficult to convey
<joepie91> focus on it in public communication, I mean
<joepie91> solving the minor problems can defeat that approach
<energizer> i think the 'automatic inspections' use case seems pretty weak. just put a camera on the wall
<samueldr> eh
<samueldr> if you say so
<samueldr> I won't try to change your mind
<energizer> do you think a robot is cheaper or more efficient than a camera? i've never been on an oil rig so i'm not familiar with that environment
<samueldr> I don't know
<samueldr> am I accidentally a sales rep now?
<samueldr> I listed use cases I read on, IIRC, an arstechnica article
eyJhb has joined #nixos-chat
eyJhb has joined #nixos-chat
<energizer> i just meant to chat, not to put you in any kinda pigeonhole :)
<energizer> but if you wanna sell me one at a hefty discount i wouldnt be opposed...
<samueldr> haha, I can sell you one 50% off my regular price
<samueldr> (I resell them with 300% markup though)
<energizer> cant wait for nixos-on-robot
Baughn has quit [Ping timeout: 240 seconds]
Baughn has joined #nixos-chat
ElXreno[m] has joined #nixos-chat
<srk> or put a camera on a quadcopter for fraction of the quadruped cost .. :)
<ElXreno[m]> Sorry for noob questions here, but I need help with the nvidia driver.
<ElXreno[m]> I have a laptop with nvidia card which supports only 390xx proprietary driver. After installing and rebooting I see a working gdm and I can log in, but after log in Xorg didn't see any displays and uses the virtual display. How I can fix this?
<srk> try #nixos
<ElXreno[m]> <srk "try #nixos"> There is silence in #nixos, so I decided to ask here.
<srk> nixos support questions are offtopic here :)
<ElXreno[m]> Oh, ok, sorry
<srk> try asking again later or discourse for better persistence
<supersandro2000> I want some feedback on my new idea: I want to run bloaty (https://github.com/google/bloaty) on all binaries build by nixpkgs-review to check, if we still have debug infos in binaries
<supersandro2000> Would you think this is useful?
<supersandro2000> I would filter the ouput for debug_info
<supersandro2000> or debug
<srk> objdump might do
<andi-> supersandro2000: I am seriously not sure why you put so much effort into *building* things when we already have ofBorg doing that for us. nixpkgs-review always appeared as a tool to manually *test* the binaries in addition to our common CI. I am honestly a bit annoyed by seeing nixpkgs-review comments in PRs that provide almost no additional benefit over ofBorg. We should improve the generic CI
<andi-> on all PRs and not some tooling that might be ran against a PR.
<bbigras> I was also wondering if automation was doing the same thing as ofBorg.
<supersandro2000> ofborg does not build dependant packages, checks are not red if they fail and are super easily missed and I can't easily hook extra things into it. Darwin is also always overloaded. Changed packages are always a few clicks away and failed builds are not shown by default.
<supersandro2000> also it would take way longer than I want to wait to get new experimental features into ofborg. See it like a test for new things we might want to integrate later in the default CI.
<energizer> shouldn't that situation be changed? how much does hydra's compute cost?
<supersandro2000> or long would it take until ofborg checks binarys with bloaty and makes the results visible? Also ofborg is hard to test.
<supersandro2000> energizer: I am not replacing hydra. if a PR has more than a 100 packages I am skipping it.
<energizer> or, how much does ofBorg's compute cost?
<supersandro2000> also I already catched way more errors than I liked in wrong dependencies in python because PRs are not tested good enough
<samueldr> that discussion should have been on #nixos-dev :), including the initial question from supersandro2000
<samueldr> (many nixos devs are not idling here, it's unfair to them to discuss here)
<supersandro2000> also when ofborg fails and I have no clue why I would need to locally build the same PR making the interaction time higher than I like which would mean I would just skip the debugging
cosimone has joined #nixos-chat
<supersandro2000> samueldr: It is not an official nixos topic so I did not want to notice everyone. More a personal effort to test new ways and things before making them official
<samueldr> it is discussion about nixos development
<samueldr> not everything on #nixos-dev is about official tools
<colemickens> gchristensen: I know you had docbook.rocks, can I ask your thoughts on asciidoc and Antora? I like asciidoc/docbook's xrefs, code-block callouts, etc, and am looking for a slightly more flexible+themable asciidoc-powered mdbook. Antora looks neat, but also like a lot going on for a project I'd not hear of before.
<samueldr> colemickens: I'm not graham, but I can say I like using asciidoc(tor) to write docs for Mobile NixOS
<samueldr> the site is all authored using asciidoctor
<colemickens> oh really.
<colemickens> nice
<colemickens> that's excellent to know
<samueldr> though it's using some messy stuff around asciidoctor for the site generation
<colemickens> aha!
<colemickens> I was looking at a2x -> pandoc -> html? But I have no idea. Need to switch gears to authoring content soon though :P
<samueldr> with that said, it's malleable enough that I never hit a problem where I couldn't do a thing
<samueldr> colemickens: yikes, looking at the adbook example output and it's... not good
<samueldr> (the demo in terminak dark theme)
<samueldr> I mean, design-wise
<samueldr> I also don't understand why it scrolls horizontally
lunc has quit [Ping timeout: 265 seconds]
cosimone has quit [Quit: cosimone]
<__monty__> Feels like some of the JS presentation softwares.
<colemickens> samueldr: saw that last night but it seems like a very fresh repo so I was reserving too-harsh-of judgement :)
<colemickens> oh its powered by asciidoctor, clearly I need to look more at asciidoctor
lunc has joined #nixos-chat
<colemickens> the horizontal scroll is particularly kind of funny as it seems rather intentional: https://github.com/toyboot4e/adbook/issues/1
<{^_^}> toyboot4e/adbook#1 (by toyboot4e, 9 weeks ago, open): Scrolling the sidebar
<samueldr> I can't seem to find how one would *actually* navigate the book
<samueldr> it seems horizontal scrolling is required to make sense of the content
<samueldr> but yeah, maybe teething troubles?
<supersandro2000> finding the right channel is not always easy for me
<colemickens> I'd really like some more SSG-like things too, that I've not seen examples of with any existing docbook/asciidoc tooling. I'd like to be able to "tag" articles and then offer static pages of articles listed by tag, and also maybe ideally a static-site-friendly js-powered search to allow filtering based on multiple tags. I feel like I've seen these features in zola/hugo type tools.
<samueldr> colemickens: ssg?
<colemickens> static site generator
<samueldr> colemickens: the site-map here is not a feature from asciidoctor https://mobile.nixos.org/sitemap.html
<samueldr> it's added (badly) on top of it
<samueldr> and from my use I think your desire for tags could work
<colemickens> hm
<samueldr> you can get *all* pages for the site, and do *stuff*, frontmatter is available
<samueldr> I use the frontmatter in blog posts to configure some things like the post image
<samueldr> blog posts -> project news :)
<colemickens> Did you look at Antora at all?
<samueldr> not really
<samueldr> I saw it would increase the closure size by including another language's ecosystem and didn't look further
<samueldr> also learned of it *after* I was done making the site
<samueldr> so there's me not wanting to re-do everything again
<samueldr> looking again, it seems to do way more than I want
<samueldr> >> Manage versions like a pro
<samueldr> >> Remote and local, private and public, bare and non-bare
<samueldr> it seems to be used in a way where you use it to host the docs, versioned
<samueldr> (which can be perfectly fin)
<samueldr> fine*
<samueldr> I don't know how to express it... but it seems to also be a *publishing platform*?
<colemickens> yes.
<colemickens> It looks like there's good stuff sandwiched in between a lot of stuff that seems... like a framework trying to manage too much for me.
<samueldr> and IIRC at the time (about a year ago?) I wasn't really fond of the output of their manual
<colemickens> oh, I really like it. this is a random page of someone that modified the antora default ui: https://developer.lightbend.com/docs/akka-platform-guide/microservices-tutorial/complete-entity.html#_unit_test
<colemickens> but of course that is subjective :)
<samueldr> things might have changed a lot in a year :)
<samueldr> I don't dislike the current output, so maybe things evolved
<colemickens> as a beginner to asciidoc with no experience with asciidoctor, I like the premise that antora and the UI are separate. I feel like with a handlebars template, I could achieve a lot of what I want with all of the frontmatter available to me (and it looks like asciidoc has 'keywords' as a first-class feature)
<samueldr> as a web integrator, I seem to remember that it was hurting me to use the docs, but not right now
<samueldr> well, found a nitpicking issue, but that's, like, because everyone who does web design and integration are using macs or systems without a tangible scrollbar
<samueldr> https://docs.antora.org/antora/2.3/ <- in a narrow window so it uses the "mobile" version, show and hide the side bar, the scrollbar will come and go
<samueldr> though, a good thing is they didn't fall for the easy pitfall of making the side bar actually disappear so widening the window again means your sidebar is lost
* colemickens is so happy every time he can join a gitter room from matrix
cole-h has joined #nixos-chat
<colemickens> samueldr: is hte mobile-nixos website source availalbe? I don't see it in the gh org
<samueldr> yes
* samueldr waits on browser that decided to take ages to render a page
<samueldr> oh, the whole computer is lagging
<samueldr> the implementation is here: https://github.com/NixOS/mobile-nixos/tree/master/doc
<samueldr> except for the news parts, and "actual hosting", which is done here: https://github.com/NixOS/mobile-nixos-website
<samueldr> it was my opinion that news aren't documentation
<colemickens> oh you did write ruby
<samueldr> yes
<samueldr> for me it wasn't a deal breaker, I'm not averse to ruby
<colemickens> I just wasn't seeing how some of this was possible with just the builtin templating support
<energizer> if you prefer python there's https://pypi.org/project/asciidoc3/ and of course sphinx
<samueldr> yeah, sorry if I misled you
<colemickens> It seems like it asciidoctor is missing any sort of extensibility that lives outside the context of a single article? and that's where you wrote the ruby bits - sitemap,rss, you had to build up those structures yourself
<samueldr> I did say "badly" :)
<colemickens> samueldr: not at all, I'm just completing the picture in my head
<samueldr> though what's great is that everything is laid bare
<colemickens> (mostly because some of this seems obvious to build into asciidoctor?)
<samueldr> nothing in private internal structures you can't look into
<samueldr> I guess maybe into an actual extension
<__monty__> Since you mentioned pandoc. Isn't that an option? You can write filters in several languages, all languages if you don't mind marshalling to and from JSON.
<samueldr> but I didn't want to get started with rubygems, while we currently have an asciidoctor package that works
<colemickens> that's what I mean though, from all appearances the extension documentation only talks about acting on the context of a single article
<colemickens> but anyway, I'm just spitballing now, this is all super useful info.
<samueldr> hm, it's so far back that I don't remember the details
<colemickens> and yeah __monty__ I'll probably poke at that a bit more too
<colemickens> oh no, when I google "a2x pandoc" now I get the issue I just filed on conix.
<colemickens> which is another idea that I had/have: https://github.com/theNerd247/conix/issues/34
<{^_^}> theNerd247/conix#34 (by colemickens, 13 hours ago, open): Use a2x and pandoc to support asciidoc?
<__monty__> What's a2x? Doesn't pandoc support asciidoc for reading?
<energizer> if conix files can get correct syntax highlighting i'm gonna be quite impressed
<colemickens> __monty__: not from what I saw. I guess maybe a2x was wrong, maybe just asciidoc is used to convert asciidoc to docbook. still learning here :P
<colemickens> energizer: if it's all a DSL, couldn't you write an expression to run a file through a colorizer? (I don't know, just speculating on what I thought conix's value prop was, and seems possible given the way it uses pandoc already)
<__monty__> Ah, asciidoc's only supported as output format yeah.
<colemickens> (while I'm falling down this hole, there's also nix-styx to poke at too)
<energizer> colemickens: the complication is dynamically figuring out what language it is. a lot of tools sorta give up on that
<colemickens> uh nix-styx has asciidoc support :O
<colemickens> energizer: oh I'm used to having to tell them anyway I guess
<__monty__> Oh, one thing. Since you specifically mention some asciidoc features, like callouts. I'm not sure pandoc conserves these during conversion.
<energizer> colemickens: i mean because a conix file has multiple languages in it
<energizer> nix and html/markdown
<colemickens> but the conix file is just nix using expressions to generate those things
<colemickens> afaict
<colemickens> __monty__: yeah, that's the first thing I want to check
<energizer> colemickens: yeah that's just what i'd call giving up :)
<energizer> that block should be highlighted in dot/graphviz syntax colors
<__monty__> Since it's tagged it wouldn't be impossible for a syntax highlighter that's smart enough. Like markdown's ```bash\ncode block syntax\n```
<energizer> yep it's not impossible
<colemickens> styx has some neato stuff in it
* colemickens wonders how styx holds up in SSG speed tests :P
<energizer> i worry that some of these systems aren't sufficiently engineered to avoid injection issues
<energizer> like, is that just string interpolation
lunc has quit [Ping timeout: 240 seconds]
<__monty__> Does it matter much? It's not like you're rendering user input.
__monty__ has quit [Quit: leaving]
<energizer> it'll be incorrect; that might matter or might not depending on what goes in there
<samueldr> depending on who authors what, it could end up at best breaking the site, at worst something nefarious
<samueldr> it's less of an issue than on a dynamic website, but still problematic
<energizer> strictly speaking the best is it just works fine. that's probably the most common case in practice tbh
<samueldr> sorry, best case when it causes issues
<energizer> mhmm
ece3 has quit [Quit: The Lounge - https://thelounge.chat]