endformationage has quit [Ping timeout: 244 seconds]
<gchristensen>
das_j: I tried looking for ajs124's repo with OSPF & wireguard but couldn't find it. any links?
<samueldr>
pie_: something doesn't seem right with my tests... the source linked doesn't match with what I observe
<samueldr>
oh, no, just understood
<infinisil>
Wow massive Google outage today
<infinisil>
It's funny to see what all breaks down because so many things rely on Google
<infinisil>
Apparently ~4 hours outage..
<elvishjerricco>
Is there a way to see what syscall a process is currently blocked on? My bluetoothd is currently uninterruptible (I think that's the term), and I wanna know what syscall it's on. strace obviously doesn't output anything since it's not making any *new* syscalls
<Ralith>
attach gdb?
<elvishjerricco>
Ralith: gdb just hangs after "Attaching to process 20552" :P
<Ralith>
aw
<elvishjerricco>
Couldn't even ctrl-c gdb. Had to kill -9 it.
<gchristensen>
elvishjerricco: look in /proc/<pid>/? I think one of those fileshas it
<eyJhb>
srhb you should have warned me about this ... humid, warm, weather. In Aalborg, it's unbearabel :(
<Shados>
19C 80% humidity... you should try Queensland, Australia, or Fiji some time xD.
<adisbladis>
Shados: Those places got nothing on south china when it comes to humidity ;)
<lucus16>
My derivation for the gnat compiler is unstable :(
<lucus16>
Sometimes it fails and sometimes it builds succesfully
<das_j>
gchristensen: What ajs124 said. If you need any details on configuration, just ask :)
<colemickens>
"--ephemeral-os-disk : (Preview) Allows you to create an OS disk directly on the host node, providing local disk performance and faster VM/VMSS reimage time. Allowed values: false, true."
<colemickens>
:D
<eyJhb>
Shados I feel like that is going to be a big, no from here, :p Just hate the humidity... :/
<srhb>
eyJhb: Sorry! We have, like, a whole website dedicated to this kind of thing.
<srhb>
Unpopular as it may be xD
<eyJhb>
srhb but I feel like there needs to be a integration with IRC, so that I don't need to open it :p
<srhb>
eyJhb: You'll be happy to know that an API is in its (very) early stages. ;-)
<eyJhb>
srhb HTTP REST API? :D :D
<eyJhb>
pie_ btw. I am not happy with the solution for RCE...
<gchristensen>
mine is a bit sloppy because my ~ is very stateful
drakonis_ has joined #nixos-chat
<infinisil>
I think everybody's ~ is like this
<infinisil>
Having that declarative takes quite some work
<samueldr>
working on it by patching software or wrapping software :)
<pie_>
cant remember if it was maybe michaelraskin that did some stuff, but he isnt on right now
drakonis has quit [Ping timeout: 258 seconds]
<gchristensen>
shlevy's ~ is r/o
<infinisil>
Nice
<infinisil>
Though that probably makes using new software were annoying
<pie_>
"you want a config location? no. shoo."
<pie_>
"get off my lawn"
<eyJhb>
Cue all the errors in the world, switching out my storage backend :/
<eyJhb>
pie_++
<{^_^}>
pie_'s karma got increased to 2
<eyJhb>
Can I actually stack my negative karma?
<eyJhb>
eyJhb++
<{^_^}>
eyJhb's karma got decreased to -2
<eyJhb>
I guess I can
<gchristensen>
aszlig++
<{^_^}>
aszlig's karma got increased to -666
<pie_>
w
<pie_>
what
<eyJhb>
I feel like that is hard fixed to -666? :%
<pie_>
:D
<eyJhb>
aszlig++
<{^_^}>
aszlig's karma got increased to -665
<pie_>
damnit!
<eyJhb>
I guess not
<pie_>
aszlig--
<eyJhb>
gchristensen is just the devil I guess
<eyJhb>
Conjuring 666's out of nowhere :p
<gchristensen>
heh
<eyJhb>
The weirdst thing about the move from my side is, that I am currently removing SOOOO MUCH code...
<Shados>
Kim: I just* got around to looking at your vim/SyntaxRange Nix expressions again, and I've finally remembered that Nix has block/inline comments (`/* stuff */`). Maybe you could use an inline comment preceding the embedded string to avoid the need for the `ft` attrset (e.g. start a def with `stuff = /*ft=vim*/''`)? *: I went mad a couple of weeks ago and decided to write a standalone module for configuring neovim, which was distracting.
ivan has joined #nixos-chat
<adisbladis>
infinisil: My ~ is not stateful at all (except for a few opt-ins)
<gchristensen>
how did you do that, adisbladis?
<adisbladis>
gchristensen: It's on tmpfs and using home-manager
<gchristensen>
good grief
<gchristensen>
I don't have the vigilence to not put important stuff in the tmpfs part :P it'd _have_ to be r/o or I'd lose all sorts of stuff :D
<adisbladis>
If I lose something it was obviously not important enough to save it in the correct place ;)
<gchristensen>
you should see the crisis which is my ~
<gchristensen>
(I'm impressed)
<adisbladis>
gchristensen: Actually I _love_ that I can just quickly hack on something in ~/ and not bother with cleaning things up :)
<pie_>
HOME=$(mktemp -d)
<eyJhb>
rm -rf $HOME
<pie_>
unset HOME
<Shados>
adisbladis: I think if I put mine on tmpfs I'd end up chucking something overly large in it and OOM myself, heh
<adisbladis>
Shados: Happened a few times :P
<adisbladis>
That's the biggest annoyance by far
<Shados>
Instead, /tmp on my zfs systems is actually a zfs dataset with `sync=disabled` and `boot.cleanTmpDir`, for tmpfs-like behaviour
<Shados>
Get most of the benefits with that approach
<eyJhb>
OOM?
<eyJhb>
Ah
<eyJhb>
In the meantime, my home folder is 117GB
<eyJhb>
And I do have a secoundary SSD for storage...
endformationage has joined #nixos-chat
<pie_>
hot take: nixpkgs is a giant expert system
<samueldr>
hot?
<pie_>
s/giant//
* pie_
rubs hands together
<pie_>
we should get it more expert system-y
<pie_>
whatever that means
<joepie91[m]>
pie_: isn't that basically what the application integration stuff does? :)
<Taneb>
I've heard that before, either a blog post or maybe a tweet or at NixCon last year?
<pie_>
i may have heard the idea before, with respect to modules, now that i think about it
<pie_>
but anyway yeah i was just thinking about how we have an (actually usable at least) language for describing domain specific information, and we are able to collaborate on it, so we can make steady progress in the forward direction hopefully without repeating a lot of mistakes
<pie_>
and just kind of thinking about the knowledge we have encoded
<pie_>
man i need to go read one of those expert systems textbooks at some point, sigh :P
<pie_>
it just bothers me when experts (or people in general) work on things and the results arent transferrable, or something like that
<pie_>
>/rant rant rant and stream-of-consciousness>
<pie_>
preaching to the choir
<Taneb>
Seems we need some sort of organization rules for these people with domain knowledge
<Taneb>
A kind of expert system, if you will
<pie_>
;D
<pie_>
computers are force multipliers if you use them right
<samueldr>
well, 0.5 is a multiplier
<pie_>
which we at least seem to be going in _a_ correct general direction \o/
<gchristensen>
lol
<pie_>
samueldr, oog
<pie_>
* oof
<pie_>
so high?
<samueldr>
you were saying "if you use them right"... you can use them wrong and still have a force multiplier :) though one that reduces whatever you multiply
<joepie91[m]>
pie_: I can only agree with all of that, really :P
<eyJhb>
*looking up expert system* - nixpkgs is AI.
<pie_>
eyJhb, oh dear
<pie_>
he said too much get the van
<eyJhb>
Well, you would pretty much need to make a roadtrip to get me... :p
<eyJhb>
I.. I feel something is very very wrong. Reduced my code from 5.000 lines to 2.500. SOmething seems waaaay off
<eyJhb>
Ahh, that seems right. From 5.000 to 6.000..
<Taneb>
That's a reduction of -1000 :D
<pie_>
niiice
<eyJhb>
Precisely! Didn't really know if what I did would increase or decrease it, since it removed a lot of logic and moved it to a outside package.. But actually only really increase by around... 250 lines :p And made it much easier to extend :p
<eyJhb>
Now I just have to test it, and that doesn't seem like fun.. Or write test cases, which seems like even less fun :(
<adisbladis>
The whole bronzy thing is encapsulating the DD
<andi->
are they comfy to wear?
<adisbladis>
Yeah I can wear them all day
<adisbladis>
I can forget that they are there
<andi->
I am still looking for ones that also come with a wireless (maybe not bluetooth) solution for my every day listening
<adisbladis>
andi-: Though I cannot recommend my ones for everyone :P
<adisbladis>
For starters you'll have to go to asia
<adisbladis>
They only do them in a few asian cities
<adisbladis>
And then you'll pay a price that makes you wanna cry
<adisbladis>
And then you wait for almost three months
<andi->
If you care about proper sound and a good experience it is totally worth it
<adisbladis>
Totally, just very expensive even for CIEMs
<adisbladis>
~2kilobucks
<andi->
Calculate the price per day/year/month/… you'll be using them..
<andi->
I used to wear mine ~8-10h a day
<adisbladis>
Me too :)
<adisbladis>
Except when I'm home
<andi->
just to get away from all the distracting noises around me and have some good audio
<adisbladis>
andi-: Planning to geat UEs again?
<andi->
adisbladis: not sure..
<andi->
I haven't decided
<andi->
haven't even started researching
<adisbladis>
I didn't like their sound sig :/
<cransom>
i bought some nice shures that i only slightly regret. they sound really nice, but i realized that my hearing is not as great as it was 10 years ago and any old driver now is probably fine. at least, when i crank up the 17khz tone over speakers, i can't really notice, but my partner yells at me.
<adisbladis>
cransom: Which ones?
<cransom>
se... 535? i think
<adisbladis>
I've had a few Shure IEMs over the years
<adisbladis>
cransom: I can highly recommend modding them :)
<andi->
adisbladis: I fear that reading up on all the changes in the last years will be a huge distractions for weeks to come.. So I just skipped it for now
<adisbladis>
I was not happy at all with my 535s until I modded them
<cransom>
what does one do to mod iems?
<adisbladis>
andi-: Seems sensible.
<adisbladis>
andi-: I waited until I was back in asia.. Much easier to try things out when you have headphone shops with ~100s of models ready to try
<adisbladis>
cransom: In this case change the acoustic filters
<andi->
adisbladis: yeah, are you using some kind of alternativ DAC/AMP/… on the go?
<adisbladis>
andi-: I'm using the Sony MUC-M2BT1 (bt w/ mmcx)
<adisbladis>
Also apples usb-c adapters are fantastic and cheap
<andi->
that makes the Sonys a bit more interesting
<adisbladis>
andi-: That's just the cable though, you can use any MMCX headphones with them
<adisbladis>
cransom: I _think_ I liked the green filter the most. That tuning is like the SE535 LTD
<adisbladis>
It really adds some of that missing treble extension to the 535s
<adisbladis>
Turns it it was the grey ones :)
<andi->
UE changed the connector system (again?)... Mine used to get loose sometimes during the day.
<adisbladis>
Oh yeah they are using that proprietary stuff nowadays..
<adisbladis>
IPX?
<andi->
yeah :/
<andi->
so maybe something else then
<adisbladis>
I'd wanna try the Kumitate KL-REF again
* adisbladis
is a sucker for DD
<cransom>
when shipping is more than the item... but oh well. lets try out some brown dampers.
<adisbladis>
andi-: I even tried contacting UE about getting connectors and they said no :/
<andi->
adisbladis: My last experience with the customer service was no reply
<adisbladis>
cransom: Word of warning, when you remove the old dampers they might get ruined
<andi->
I send them a few reminders and called
<andi->
no reasonable reply
<adisbladis>
=(
<adisbladis>
I usually hear good things about UE support
<cransom>
yeah, i saw that cautionary tale. i took it as a challenge to not destroy them.
<adisbladis>
cransom: It might be impossible
<adisbladis>
Iirc my ones were glued on
<adisbladis>
Not that I think you'll regret changing them =)
<cransom>
at worst, i'll just only listen to them over my bluetooth adapter and blame that for why i hate them.
<adisbladis>
:D
* joepie91[m]
is disliking the `with` keyword in Nix more and more
<joepie91[m]>
makes it so difficult to figure out where stuff comes from...
<adisbladis>
cransom: Buy the Shure bluetooth adapter and you'll hate it for shure ;)
<adisbladis>
joepie91[m]:++
<samueldr>
joepie91[m]++
<{^_^}>
joepie91[m]'s karma got increased to 1
<joepie91[m]>
I'm half-tempted to go on a with-removal spree in nixpkgs, and just open PRs that nukes it out of code and replaces it with explicit referenes
<joepie91[m]>
but something tells me that might end up being controversial :)
<samueldr>
I only use it in ultra-local places like `with pkgs; []` and much rather `let inherit (lib) func func func; in`
<gchristensen>
stm joepie91[m]
<joepie91[m]>
gchristensen: stm?
<gchristensen>
+7
<gchristensen>
I've forgotten how to type.
<gchristensen>
sounds good to me
<joepie91[m]>
ah :)
<samueldr>
removing those "global" one at the outermost or next-to-outermost scope I feel wouldn't be met with much pushback
<samueldr>
but maybe some for those in meta for derivations
<joepie91[m]>
samueldr: I expect most pushback on the lib ones
<joepie91[m]>
(which incidentally are the very same ones that usually drive my confusion, alongside pkgs)
<samueldr>
`with lib;` in an outermost scope? please make them go away and `let inherit` them
<samueldr>
(my opinion!)
<joepie91[m]>
yes, I've run into `with lib`s in a few places
<joepie91[m]>
don't recall where
<samueldr>
I did too, just wanted to be sure that `license = with lib; [ license.x license.y ]` weren't necessarily in sight :)
<pie_>
Church-, ping me when the repo is up ;=
<pie_>
* ;)
<pie_>
same <samueldr> I only use it in ultra-local places like `with pkgs; []` and much rather `let inherit (lib) func func func; in`
<pie_>
i think with is nice to have sometimes but a lot of removals could probably be justified
* pie_
is still waiting for the better debug messages for errors i strings tho :(
<joepie91[m]>
I'm increasingly convinced that `with` should be nerfed in Nix like it is in JS
<joepie91[m]>
as in, on a language level
<pie_>
meh, we should switch to scheme :P
<joepie91[m]>
technically there are some valid uses, but it looks too attractive too often in the wrong scenarios with too-significant readability issues as a result, imo
<pie_>
keeping the purely functional semantics
<joepie91[m]>
to justify its continued existence
<pie_>
joepie91[m], you should peek at my repo first ;PP
<gchristensen>
that one was real flat. I loved that one's design
<samueldr>
right, for the display the holes are a big phat heatsink
<samueldr>
starts at 5999$
<samueldr>
also available as a rackmountable version
<gchristensen>
look out macstadium, don't need your fancy rack adapters anymore
<samueldr>
those new displays are 4999$ or 5999$ (1k for matte display) that doesn't include any mounting
<samueldr>
199$ for vesa, 999$ for their mount :/
<gchristensen>
lol.
<samueldr>
though, they *are* 6K displays
<gchristensen>
how many Ks do we need anyway?
<samueldr>
apparently 6
<samueldr>
sorry, 6 6K
<samueldr>
since the mac pro will handle 6 6K displays
<gchristensen>
I suppose that is adequate
<gchristensen>
I am really grateful that Apple has pushed hidpi to where it is
<samueldr>
GRATEful... now that's just on the nose :)
<gchristensen>
hehe
<lucus16>
:')
<joepie91[m]>
thoughts on replacing the usage of builtins.trace for deprecations (like in https://github.com/NixOS/nixpkgs/blob/master/lib/trivial.nix#L162) with a dedicated 'deprecated' function? that functionally does the same (a trace), but would allow eg. future integration of deprecationg tracking tooling, by explicitly flagging deprecated things in a machine-interpretable format
<gchristensen>
sounds neat
<samueldr>
errrr... iOS and macOS devices will broadcast bluetooth beacons that other iOS and macOS devices will relay using their network connection for offline location tracking for "find my"
<gchristensen>
cool cool cool
<__monty__>
Sounds like we might get free network traffic if we play our cards right.
<joepie91[m]>
gchristensen: also, do you think that's something that should be defined on a nixpkgs level, or on a language level? I'm tending towards 'language level' given that it isn't nixpkgs-specific at all, but OTOH that might open the door to a lot of "why don't we have this utility in the language as well", and I'm failing to come up with a clear delineation that would prevent that and still support this being in core
<gchristensen>
joepie91[m]: nixpkgs level to start :)
<joepie91[m]>
gchristensen: any particular rationale?
<gchristensen>
languages are slotw
<gchristensen>
slow*
<gchristensen>
language features are permanent
<joepie91[m]>
as in, the process to add stuff?
<gchristensen>
yeah
<joepie91[m]>
right
<gchristensen>
like >1yr before nixpkgs can use something in a new nix
<joepie91[m]>
in hindsight, that is an obvious rationale, not sure why I didn't look at it from that angle :D
<joepie91[m]>
thanks lol
<gchristensen>
:)
<gchristensen>
later on we can always do lib.whatever = builtins.whatever
<joepie91[m]>
made worse by the fact that I frequently explain this rationale to people in the context of JS
<joepie91[m]>
I guess I somehow didn't mentally associate it with Nix
<gchristensen>
haha
<gchristensen>
samueldr: I'm annoyed by all the people comparing the mac pro to the least similar cheese grater.
<samueldr>
hm?
<samueldr>
like they're not even taking time to find one that looks right?
<gchristensen>
yeah
<joepie91[m]>
hold on
<joepie91[m]>
I can fix that
<samueldr>
be the change you want to see in the world
<samueldr>
make yer own shitpost :3
<joepie91[m]>
gchristensen: ah, the handle is a little different now than it used to be (it used to be more similar!) but https://www.action.com/nl-nl/p/rasp-/
<andi->
Well nix has some different rules (internally) for diffing versions.. if the string `pre` exists or some other string it does it differently
<joepie91[m]>
andi-: I mean, if you can tell me wtf that is for... :D
<andi->
joepie91[m]: no idea :D
<andi->
I mean yes I can see the usage and maybe it comes in handy somewhere.. I am not seeing the bigger picture/pattern for nixpkgs..
<joepie91[m]>
it's used in nixos/doc/manual/default.nix and nowhere else it seems
<joepie91[m]>
and I still have no idea what it's actually doing there!
qyliss has joined #nixos-chat
qyliss^work has joined #nixos-chat
<gchristensen>
it is splitting the type in to two subtypes by predicate p, takes all the elements of the first subtype to be less than all the elements of the secondsubtype, compares elements of the single subtype with yes and no respectively. isn't it clear? trivial!
<lucus16>
I feel like wikipedia and the archwiki look much more professional and that causes them to inspire more professional content
<samueldr>
mediawiki doesn't have a "new page" button
<lucus16>
I guess
qyliss has joined #nixos-chat
<lucus16>
Well, yeah, that isn't the primary issue, it just doesn't feel like a wiki
qyliss^work has joined #nixos-chat
<gchristensen>
what would make it feel more wiki-ey?
<lucus16>
wikipedia/arch wiki lookalike, basically the mediawiki look
<__monty__>
Isn't the mediawiki look based on wikipedia?
<samueldr>
monobook is... it has its limitations
<samueldr>
(monobook being "that wikipedia look")
<lucus16>
<__monty__>: Exactly
<__monty__>
But how can wikipedia look more professional if it's the same look?
<lucus16>
I mean tbh, I like the cleaned up look of the current wiki much more in general, but I always forget its a wiki and that I can contribute improvements because of it and I fear that goes for others too
<samueldr>
__monty__: they (archlinux wiki, wikipedia) compared to nixos.wiki
<lucus16>
^
<samueldr>
lucus16: there is one thing that you're entirely right on target: no CTA (call to action) when logged-out
<lucus16>
The edit link is a tiny one that looks like it's just part of the footer
<samueldr>
no obvious "hey look maybe you could edit me" button
<lucus16>
And I'm trained to ignore anything in footers
<samueldr>
you're right, in web design the footer "is invisible"
<samueldr>
it should still exist and provide functions, but not be the only location for them ideally
Kim has joined #nixos-chat
<__monty__>
I agree the edit link could be more visible. But the '.wiki' tld should tell most everyone it's a wiki.
<samueldr>
(btw, I integrated the theme on nixos.wiki, so I'm not talking out of my ass :))
<lucus16>
__monty__: If you notice it, sure
<samueldr>
__monty__: annoyingly, the theme used has a more obvious button... when logged-in
<lucus16>
Most of the time I just get there from google and I never notice the URL
<lucus16>
To me, the part that reminds me most of all that something is a wiki is the left sidebar
<samueldr>
main issue, right now, is that contributing infra changes for the nixos.wiki is high in friction, I have a couple changes in the pipeline I haven't tackled
<lucus16>
And to some degree the top one
<lucus16>
Because all good wikis have that
<lucus16>
I mean no offense, but the nixos wiki is severely lacking content when compared to the Arch wiki
<lucus16>
And I notice myself constantly looking for things I've solved before but forgotten again
<samueldr>
the age of the wiki, and the size of communities have to be taken in account for
<lucus16>
And a wiki is a perfect place to put those
<lucus16>
Like the zip 1980 thing
<lucus16>
factoids are fun but they're not googleable, they only really work when someone is around that remembers the relevant factoid
<joepie91[m]>
from the docs for `.toJSON`:
<joepie91[m]>
> Derivations are translated to a JSON string containing the derivation’s output path.
<{^_^}>
error: syntax error, unexpected $undefined, expecting ')', at (string):254:70
<joepie91[m]>
does that mean that they are evaluated?
kgz has quit [Ping timeout: 248 seconds]
<joepie91[m]>
or does it just obtain the output path it would have?
<__monty__>
lucus16: The archwiki is a very rare example of a good wiki. So not exactly a fair comparison.
<lucus16>
I know that, but I feel we can do better
<__monty__>
Also, much of the purpose the arch wiki serves is already spectacularly fulfilled by the nix(pkgs) manuals.
<lucus16>
And I feel the wiki is obscure because if other people are anything like me, they won't recognize it as a wiki
<lucus16>
better than currently, not better than arch mind you
<lucus16>
The manuals are quite a bit less searchable imo
<lucus16>
I couldn't find the 1980 zip hook in the manuals
kgz has joined #nixos-chat
<lucus16>
I really want a page comparing the differences and usages of all the different fetchgits
<__monty__>
They're pretty searchable for me. Do you use google?
<lucus16>
yes
<lucus16>
If I google the exact error message, "ValueError: ZIP does not support timestamps before 1980", the manuals don't show up on the first page
<__monty__>
¯\_(ツ)_/¯ I guess no one's stopping you from creating the pages you'd like to see : )
* joepie91[m]
just so happens to be writing "I want to do X in Nix, how do I" docs at the moment
<lucus16>
__monty__: I intend to
<joepie91[m]>
hence above toJSON question, which I hope somebody has the answer to :)
<samueldr>
(just a note, a long time nixpkgs user's google results are different than a non-nixpkgs user, highly annoyingly)
<joepie91[m]>
also, mark me down as another person in the "nix manual is very unsearchable" camp
<lucus16>
But I still feel the wiki looks like cheap blog
<joepie91[m]>
but I have ranted about this before :)
<samueldr>
does nixos.org look like a cheap blog?
<__monty__>
samueldr: I don't know how much google changes its results based on fingerprinting? I never log in to a google account.
<samueldr>
__monty__: through anonymizing services?
<samueldr>
because even not logged-in it will skew I found out
<lucus16>
samueldr: imo, nixos.org doesn't look as cheap
<lucus16>
I guess it's kind of hard for me to explain what exactly puts me off, but to me, the arch wiki is much more inviting to contribute to
<__monty__>
That's why I said fingerprinting. I doubt they change the results as drastically as for someone logged in.
<__monty__>
Most contributable wiki is the github wikis imo.
<samueldr>
github wikis are the hosted service trap :/
<samueldr>
no way to http redirect them out from a github-owned URL
<gchristensen>
the github wiki is also HORRIBLE
<lucus16>
^
<lucus16>
most contributable only in terms of how easy it is to have an account
<lucus16>
Though wikipedia allows edits without account
<__monty__>
I'm not suggesting hosting on github. Just the "wiki is simply a set of markdown files in a repo" model.
<samueldr>
right, gollum model
<samueldr>
that sure is an inviting model for developers
<lucus16>
It doesn't scale that well though imo
<lucus16>
You can't easily make templates and stuff
<elvishjerricco>
So I know NixOS has some kind of issue with ZFS's automatic mountpoint stuff. I think this might be able to fix them though, right? https://github.com/zfsonlinux/zfs/pull/7329
<{^_^}>
zfsonlinux/zfs#7329 (by aerusso, 1 year ago, merged): systemd.mount integration
<__monty__>
Why not? You can just have a page per template and link to them for a "Contributing guidelines" page.
<samueldr>
nothing forbids it in that model, pretty sure gollum itself can, but for the github hosted ones you're right
<samueldr>
__monty__: templates like boxes {{warning|don't do this}}
<lucus16>
__monty__: You're suggesting copy-pasting the templates to the pages, that's not very maintainable when the templates are improved
<__monty__>
Oh, like reST admonitions?
<__monty__>
lucus16: That's true.
<samueldr>
maybe not, mostly like partials
<lucus16>
tbh, the whole mediawiki language is pretty magical for me too
<lucus16>
I don't understand much of it
<samueldr>
so you can even build complex templates with conditionals to build up data visualisations from content
<lucus16>
But I understand it was built to scale well
<__monty__>
Don't know reST partials.
* joepie91[m]
dislikes mediawiki syntax
<joepie91[m]>
(to the point that I avoid contributing)
<eyJhb>
Agree with joepie91[m] there... mediawiki syntax should die
<samueldr>
__monty__: not saying reST partials, I don't know if it's a thing, but talking about view systems partials
<lucus16>
I kind of dislike it too, but...
<samueldr>
yeah, I disliky mediawiki syntax too :/
<__monty__>
Pandoc can convert to mediawiki : >
<eyJhb>
Converted my personal wiki from mediawiki to markdown. Did not go well
<eyJhb>
And now __monty__ COMPLETELY ruined it...
<eyJhb>
Don't bring logic to this table
<lucus16>
xD
<aanderse>
ugh mediawiki...
<__monty__>
Markdown's too simple for wikis. I think reST could be a good markup for them.
<eyJhb>
Never really gotten the hang of reST...
<eyJhb>
Can you make citations and references in reST?
* aanderse
has been working on a mediawiki module to replace apacheHttpd/mediawiki.nix recently
<__monty__>
You can do everything in reST, because it's extensible : )
<aanderse>
not so much on the fun side -_-
<samueldr>
aanderse: ping me when you progress I want to test the nixos wiki config on it
* joepie91[m]
has bad experiednces with rst as well
<joepie91[m]>
experiences*
<eyJhb>
joepie91[m] do you like anything ? :p
<joepie91[m]>
markdown for the most part, though it is definitely missing a few features
<joepie91[m]>
sec
<__monty__>
Granted you may need to alter the tooling to have your extended reST work... I cheat and just use field lists for most constructs with a custom pandoc filter.
<lucus16>
I used to hate mediawiki syntax too, but after using it, it's pretty ok.
<lucus16>
It looks ugly
<lucus16>
But it's basically the most mature language for wikis
<__monty__>
Maybe org syntax'd work for a wiki?
<samueldr>
aanderse: the current configuration isn't handled via nix, not entirely sure of the details about why, but the admin isn't closed to the idea, mostly busy
<lucus16>
mediawiki is not just markup, it allows pages to affect other pages
<samueldr>
aanderse: I have the full set of requirements so I think I can test it fully, or close to fully :)
<lucus16>
E.g. if you add something to a category, that page also shows up in the category.
<joepie91[m]>
there's also another markdown derivative I ran across that extended it with more wiki/documentation-esque features, but I can't find it anymore :(
<samueldr>
joepie91[m]: actually interesting take, I really like xml-like "YOU MUST CREATE A VALID DOCUMENT" for fancy markup
<lucus16>
On a private wiki I use, we have a {{Present|Name}} which links to your profile, but also adds the page it's on to a list of events on your profile you've been to
<aanderse>
samueldr: great! i have no personal interest in mediawiki, i just want to get rid of all the httpd subservices... so testers with vested interest in valuable
<eyJhb>
lucus16 but... Markdown could do the same? It is all about parsing. But it is missing that syntax yeah
<__monty__>
Adding JS to markdown doesn't sound like it solves the issue with markdown lacking features tbh.
<samueldr>
it might
<samueldr>
depending which features
<eyJhb>
Adding JS to anything is not fun.. *looking at nodejs
<__monty__>
Well the main problem is lacking syntax.
<joepie91[m]>
__monty__: the main value is in the 'X'-y part of JSX
<lucus16>
I might kind of like racket scribble if we're talking about fancy new things
<__monty__>
Inline html is pretty unreadable sometimes.
<lucus16>
But lately I much prefer to stick to tried and proven solutions
<joepie91[m]>
also, is it JS-bashing time again?
<samueldr>
yeah, looking at it the same could be applied to any "embed xml syntax into markdown and transclude contents"
<joepie91[m]>
I'm not enjoying this becoming a recurring event
<gchristensen>
I agree with tried and proven
<gchristensen>
docbook++
<gchristensen>
joepie91[m]: I hope not, that isn't what we're about here.
<samueldr>
gchristensen: which is why I like the JSX approach of markup, same well-formedness guarantees than in XML and docbook :)
<eyJhb>
If we bash on all languages equally, isn't that fair then?
<lucus16>
I don't think we do though
<eyJhb>
I can kick haskell a little?
<gchristensen>
samueldr: nice
<gchristensen>
how about we not kick on stuff
<__monty__>
I'm not bashing JS. Just don't expect it to solve the issues I have with md.
<lucus16>
eyJhb: joepie91[m]: go join #proglangdesign, they love bashing existing languages
* eyJhb
creates #nixos-kicking
<eyJhb>
Ohh. Seems like a solution ;)
<lucus16>
Well, as long as you're also actually thinking about solutions for all the problems you're bashing
<__monty__>
joepie91[m]: You complained about reST without giving any reasons let alone examples, should I be upset about you bashing python?
<eyJhb>
The solution most of the time is not using it, and being very mad when I am forced to do so.. :/
<lucus16>
eyJhb: That's not very constructive
<gchristensen>
^
<joepie91[m]>
__monty__: I expressed dislike of the language. I'd be fine if the comment was "I don't like JS", but that's not what the comment was - instead, the comment was "adding JS to anything is not fun"
<joepie91[m]>
which is a snide remark, not an expression of personal dislike.
<lucus16>
Let's please not start this drama again
<__monty__>
*My* comment was I don't think adding JS to md solves the problems *I* have with it.
<__monty__>
I don't enjoy the idea of any programming language embedded in my markup language though. The point of these simplified markups, i.e., mark*down*s, is to be readable *as a document* first. And programming languages violate that way too easily.
<joepie91[m]>
__monty__: ah, right. to be clear, I wasn't responding to *your* comment about JS, which was not a snide remark :)
<joepie91[m]>
rather to eyJhb's
<__monty__>
Apologies then, I still have PTSD from last time : p
<gchristensen>
__monty__++
<{^_^}>
__monty__'s karma got increased to 5
<gchristensen>
joepie91[m]++
<{^_^}>
joepie91[m]'s karma got increased to 2
<joepie91[m]>
__monty__: yeah, np :)
<__monty__>
joepie91[m]: Were your issues with reST specific or general?
<joepie91[m]>
__monty__: also, I can't go into too much detail atm due to RSI, but the main thing with MDX is that it doesn't actually involve JS in the documents themselves; it'd be more accurate to think of it as a Markdown variant that can reference *externally-defined* components (which may be either just markup, or have some interactivity)
<joepie91[m]>
this addresses usecases like infoboxes, warning headers, special kinds of links, etc.
<joepie91[m]>
with a relatively standard and well-designed syntax
<joepie91[m]>
(aside from that one issue that I linked)
<__monty__>
Sounds like reST tbh.
<joepie91[m]>
the application I'm thinking of wouldn't really involve any browser-side JS at all, just JSX-as-XML-variant
<joepie91[m]>
__monty__: my issues with rst, insofar I've used it (which was a while ago), were mostly of an "I find this syntax awkward to work with" nature, similar to mediawiki syntax
<joepie91[m]>
it felt like it required too much line noise to express relatively simple things
<kraem>
kind of ot but; would you consider it unsafe to have private keys in a syncthing folder?
<joepie91[m]>
with Markdown doing a better job there
<__monty__>
Imo the basic reST syntax is almost a carbon copy of markdown, or well the other way around.
<lucus16>
kraem: Depends. Consider how easy it is to steal your private key when it's in synthings relative to when it's not
<gchristensen>
private keys for what, kraem?
<lucus16>
kraem: How do you authenticate with syncthing?
<samueldr>
joepie91[m]: know of non-js JSX parsers? lazily thinking thoughts
__monty__ has quit [Quit: leaving]
<kraem>
gpg/ssh keys
<gchristensen>
ssh keys: don't sync. make new ones
<kraem>
it's encrypted in transit right?
<eyJhb>
I feel like I missed a heavy discusion on this subject before...
tilpner has quit [Remote host closed the connection]
tilpner_ has joined #nixos-chat
<lucus16>
kraem: Yes, but with what credentials?
<kraem>
lucus16: i didn't set up anything but what comes out of the box
<lucus16>
kraem: If it's just with a password, then your adversary needs only to steal your password in order to access your privatekey from the internet
<lucus16>
And stealing passwords isn't particularly hard
<kraem>
lucus16: you mean password for the gpg key or the syncthing service?
<lucus16>
the syncthing service
<lucus16>
but the gpg key too I guess
<kraem>
didn't know you set up passwords for the syncthing service - thought it set up pub/priv keys behind the scenes
<lucus16>
kraem: I don't know how it works, I haven't used it, that's why I asked how you authenticate with it
<lucus16>
Do you pair devices somehow?
<manveru>
you can have a password for the web interface
<kraem>
yeah you have to authenticate with devices
<kraem>
manveru: i don't have the web interface listening to nothing else than localhost
<manveru>
yeah, just saying :)
<lucus16>
kraem: I'm unclear. How do you set two devices to have access to the same syncthing directory?
<lucus16>
Looks like you need to copy a device id, which seems to be a public key, to the other device
<kraem>
lucus16: you basically add a host identifier from machine A in machine B - machine A then gets a request to sync folders with machine B
<kraem>
yup
<lucus16>
So assuming their crypto is secure, the security mostly depends on how you moved that identifier from A to B
<kraem>
alrighty - maybe not best practice - not planning on having the keys there permanently - just while wiping my machine and installing nixOS :)
<gchristensen>
no thumb drive?
<joepie91[m]>
samueldr: "non-JS JSX parsers" are not really a thing, as JSX is by definition a superset of JS (and realistically, requires that on a pragmatic level to work)
<kraem>
need to use the only one i have for the live iso
<joepie91[m]>
samueldr: unless you mean "JSX parsers not written in JS", in which case, they might exist!
<samueldr>
joepie91[m]: I meant both
<joepie91[m]>
but I have no idea
<samueldr>
joepie91[m]: though $langX could realistically work in a bunch of languages
<lucus16>
kraem: Sounds like it is about time to set up a backup of your private keys in a secure locatino
<samueldr>
with the same kind of semantics about how it compiles down
<samueldr>
and it was more of an idle thought than actually searching for something :)
<joepie91[m]>
samueldr: the way JSX works heavily leans on functional-esque properties like lexical scoping and 'free-standing functions' and mapping etc.
<joepie91[m]>
so that rules out a pretty broad set of languages, if you want something ergonomic
<samueldr>
(a RubyX could compile down to how hyperloop manages its JSX like thing)
<joepie91[m]>
(though even in JS, it is not perfect!)
<joepie91[m]>
(mostly because we don't have a match expression yet)
drakonis has joined #nixos-chat
<kraem>
lucus16: the options i've stumbled upon are basically keeping them on paper and thumb drive in safe location - are there any other practices one should know about?
<lucus16>
Those are both acceptable options I think
<lucus16>
Much more fancy options are available too
<kraem>
I like fancy options :)
<lucus16>
You can use shamir secret sharing to split your password encrypted key into multiple shares and distribute them among trusted people and places
<lucus16>
So you can define you need 3 out of 5 shares to recover it, which means 2 people/places can be compromised and it will be safe and 2 shares can be lost and it will be recoverable
<kraem>
that sounds super interesting - added to my to-read-list :)
<lucus16>
think of close friends and family, but also a bank, a notary, a location at home, your wallet
<joepie91[m]>
odd. we have builtins.bitAnd, builtins.bitOr, builtins.bitXor... but not builtins.bitNot, for which you need to use lib.bitNot instead
<joepie91[m]>
hmmm. we don't have a fold that picks its initial accumulator from the list, it seems...
<joepie91[m]>
that seems like an oversight
<elvishjerricco>
So with the recent topic of stateless work stations, how extreme can you go with that? Someone already mentioned ro $HOME, but what about the places that *do* need to be written to, but you want to wipe it as often as possible? e.g. /tmp
markh[m] has joined #nixos-chat
drakonis has quit [Ping timeout: 258 seconds]
<elvishjerricco>
Is boot the only reasonable time to wipe /tmp?
<gchristensen>
you can setup a tmpfiles.d rule to prune /tmp
<joepie91[m]>
what. lib re-exports the builtins add, sub, and div
<joepie91[m]>
... but not mul
<joepie91[m]>
???
<joepie91[m]>
I think I'm treating this channel as my papercut log now :)
<elvishjerricco>
gchristensen: Hm but you probably can't wipe all of /tmp periodically. Some process may not be currently using any files that it expects to be there later
<elvishjerricco>
Er, that's badly worded
<elvishjerricco>
A process may be currently using no files, yet still expect a certain file to be there untouched at a later time
<andi->
give each process a dedicated /tmp folder and goes away when then process dies?
<elvishjerricco>
Good idea. It does require you to figure out how to practically containerize EVERYTHING though
<elvishjerricco>
You also have to figure out which processes should keep their parents' tmps
<joepie91[m]>
also, according to the docs, all builtins seem to be available under `builtins.*`... *except* for baseNameOf
<elvishjerricco>
I suspect "every systemd service" and "every login session" are probably good as the only places to spawn new /tmps
<samueldr>
I'd do it in multiple phases
<samueldr>
(1) identify what writes where
<samueldr>
(2) setup a /tmp for thos unidentified
<samueldr>
in (1) I would also make them write to a private-ish location
<samueldr>
or at least, a known
<elvishjerricco>
That works too. Though it can be hard to track down the processes that do promptly cleanup their tmp files, which matters if you're also going for isolation.
<samueldr>
right, though I guess with a fuse tmpfs that logs events something could be achieved%
<elvishjerricco>
Oh wow that'd be a lot :P
<samueldr>
I guess so, but maybe fewer _processes_ than files, hopefully
pie__ has joined #nixos-chat
<elvishjerricco>
Systemd actually seems to have a lot of stuff for things like this, like PrivateTmp. I need to just read the systemd man pages cover to cover :P
pie__ has quit [Ping timeout: 246 seconds]
<gchristensen>
joepie91[m]: hot dog :)
<Shados>
Train line has "major delays" and it's effectively 6C out... guess I'm working from home today
<joepie91[m]>
gchristensen: unsure if that is positive or not :D
Synthetica has quit [Quit: Connection closed for inactivity]
<gchristensen>
positive :)
<gchristensen>
Shados: where are you that it is 6C?
sir_guy_carleton has joined #nixos-chat
<gchristensen>
the credit card for my google mail account expired, and I'm debating between moving to another host, vs. updating the card. any opinions?
<Shados>
gchristensen: "effectively"/feels like due to humidity + windchill. Actual temp is like 9C. And Melbourne, Australia. We're having strangely consistent (and cold) weather at the moment.
<gchristensen>
interesting
<samueldr>
gchristensen: are there any google services on that GSuite account other than e-mail?
pie__ has joined #nixos-chat
<pie__>
i heard (?) sqlite doesnt like concurrency so...do you guys know anything else that *can* handle it?
<samueldr>
gchristensen: that's what's stopping me from moving off GSuite :/
<gchristensen>
mysql
<pie__>
thats relatively light
<gchristensen>
ehh... my wife maintains a calendar that I access through it
<pie__>
gchristensen, i kinda figured but thought i might as well ask
<Shados>
pie__: It can "handle" it, for some definitions of handle. What's your intended usage?
<gchristensen>
pie__: what kind of concurrency are you talkin'?
<samueldr>
I was more thinking about paid stuff like google play apps bought
<samueldr>
or public stuff like a youtube page
<gchristensen>
samueldr: oh .... ouch
<pie__>
Shados, nothing heavy, just ambiently wondering about possibly managing zfs metadata but i dont even have a use case
<gchristensen>
I've certainly bought _some_ apps, but I don't know how many or how much.
<pie__>
so like...a smol library catalog i guess
<gchristensen>
sqlite will probably be fine.
<samueldr>
at least I'm grandfathered into a free GSuite, BUT it's also an inconvenience as GSuite accounts can't use new google "end-user" services and free tier don't get new GSuite "businessy" services either
<Shados>
pie__: The only real problem is that you can't have multiple concurrent writers -- write locks are global. You can still have multiple writers, they just may get blocked for a little bit waiting on another writer to finish from time to time
<pie__>
Shados, i guess thats fine
<Shados>
In practice writes get queued anyway, so it's only really a concern if you have a high concurrent writer performance requirement
<pie__>
i vaguely remember some sort of rant about this
<Shados>
gchristensen: If you're looking for another email provider, I can confirm that fastmail has an *awesome* web interface, and their support is typically pretty good.
<Shados>
And of course, if you want a referral code... ;)
<gchristensen>
send one along
<gchristensen>
I might use it :)
<samueldr>
>> zsh (Z shell) is the default shell for all newly created user accounts, starting with the macOS Catalina beta, currently available only to members of the Apple Developer Program.