gchristensen changed the topic of #nixos-chat to: NixOS but much less topical || https://logs.nix.samueldr.com/nixos-chat
<pie_> qyliss might know about power stuff
cjpbirkb1 has joined #nixos-chat
cjpbirkbeck has quit [Ping timeout: 265 seconds]
Emantor has quit [Quit: ZNC - http://znc.in]
Emantor has joined #nixos-chat
spacekookie has quit [Quit: **aggressive swooshing**]
spacekookie has joined #nixos-chat
rajivr has joined #nixos-chat
<ashkitten> woah rsync is super slow wtf
<ashkitten> cp gets like 3.2GB/s copying a file whereas rsync gets only 550MB/s or so
iqubic has joined #nixos-chat
spudly[m] has quit [Quit: killed]
iqubic has quit [Remote host closed the connection]
spudly has quit [Quit: ZNC - http://znc.in]
<ashkitten> holy crap the performance of this ssd is incredible
<samueldr> what is it?
<samueldr> anything special or you had a bad one before?
<ashkitten> tried a slick rust program for copying directories recursively in a highly multithreaded fashion, and got around 2.2GiB/s on my ffxiv wineprefix
<ashkitten> the one i originally bought that i'll be returning (because it gave io errors on certain workloads, in addition to being cheap and much slower) was a kingston something
<samueldr> that makes me think
<samueldr> I should really install the SSD I god for my workstation, and set it up
<ashkitten> ssds are cool
<samueldr> on an SBC, RK3399-based, using an M.2 SSD runs around to using a (good) usb flash drive... which was already better than "usual" SD cards
<samueldr> and it's the cheapest SSD available!
<samueldr> Kingspec!
<samueldr> cheapest I can buy from a reputable source*
<ashkitten> heh
<samueldr> for a default kernel build, it went from 13 hours to 4!
<samueldr> (or maybe less than 4? I don't recall exactly)
<ashkitten> this thing is really incredible, it gets nearly 4GiB/s reads
<samueldr> but the better thing is that the machine stays responsive via ssh while building on all cores
<ashkitten> 3.4GiB/s writes
<ashkitten> and yeah that's important
<samueldr> while when it was stuck waiting for I/O, **everything** went slow
<ashkitten> been there
<samueldr> maybe one day the raspberry pi proper will have proper storage
<samueldr> (no, the rumored CM4 won't count)
spudly has joined #nixos-chat
<ashkitten> most of the reason i decided to get an ssd was not actually for the speeds at all (because zfs is performant enough for me) but because heavy writes won't interfere with anything else
<ashkitten> but it is nice to have it be this blazingly fast too
<samueldr> yep, I would have settled for SATA hdd
<samueldr> but the M.2 sata adapters boards didn't look that interesting price-wise
<samueldr> it would have made the SBC clunkier
<ashkitten> nvme drives are barely more expensive anyway
<samueldr> yeah, I mean M.2 to literal SATA plugs, with a SATA controller
<samueldr> plural plugs
<ashkitten> yeah
<samueldr> clunky since power wouldn't have been provided
<ashkitten> it's gonna feel really weird walking into kohls and handing them this stick-of-gum-sized pcba to get my refund (hey, amazon said i don't need a box or label)
<samueldr> I don't know the relation between kohls and amazon
<ashkitten> idk but apparently they do amazon return dropoffs lol
spudly has quit [Quit: killed]
spudly has joined #nixos-chat
cole-h has joined #nixos-chat
endformationage has quit [Quit: WeeChat 2.9]
waleee-cl has quit [Quit: Connection closed for inactivity]
spudly has quit [Quit: killed]
spudly has joined #nixos-chat
spudly has quit [Quit: killed]
cjpbirkb1 has quit [Ping timeout: 256 seconds]
bqv has quit [Ping timeout: 240 seconds]
iqubic has joined #nixos-chat
iqubic has quit [Ping timeout: 240 seconds]
buckley310 has joined #nixos-chat
spudly has joined #nixos-chat
da_dada has joined #nixos-chat
dadada_ has quit [Ping timeout: 272 seconds]
maxdevjs has quit [Ping timeout: 244 seconds]
da_dada has quit [Ping timeout: 240 seconds]
disasm has quit [Ping timeout: 240 seconds]
disasm has joined #nixos-chat
da_dada has joined #nixos-chat
emilazy has quit [Ping timeout: 240 seconds]
emilazy has joined #nixos-chat
cole-h has quit [Ping timeout: 260 seconds]
<sphalerite> Anyone know what might lead to a HDD "failing" (reporting lots of I/O errors, then not even responding to SMART requests), but working just fine and data being completely undamaged when placed in another machine?
<etu> sphalerite: Can be cables
<sphalerite> I'm running badblocks in write mode on it now, it's just started writing the 4th pattern, and not complained on any of the runs
<sphalerite> etu: same cable in the new machine.
<etu> Or sata controller
<sphalerite> I've heard it's very rare for a sata controller to fail
<sphalerite> though I guess that doesn't mean it can't happen :D
<MichaelRaskin> I can just say that my BRIX sometimes need reseating the SATA cable from both ends to be able to boot…
<MichaelRaskin> I guess that might mean that installing the same drive back might suddenly make it work again
<MichaelRaskin> (I also should at some point investigate what can I do with one of the two RAM slots having the contacts fall off, but all this anti-epidemic-measures stuff does not make looking for exotic soldering options attractive)
<sphalerite> this failure was a long time after boot though, so I don't think that's a likely cause?
<sphalerite> contacts falling off RAM slots? That sounds scary to fix.
<MichaelRaskin> Well, it is the place where entire slot thing gets attached to motherboard traces
<MichaelRaskin> Well, cleaning the contacts never hurts, right?
<MichaelRaskin> The contacts are in correct place, just lost their attachment (I guess from vibration during my moves, including airplane trips, two years ago?)
<MichaelRaskin> So theoretically something something hot air _might_ be enough
emilazy has quit [Ping timeout: 240 seconds]
<buckley310> I suppose it could be power supply issues
emilazy has joined #nixos-chat
buckley310 has quit [Ping timeout: 244 seconds]
buckley310 has joined #nixos-chat
sdier has quit [Ping timeout: 240 seconds]
sdier has joined #nixos-chat
<sphalerite> MichaelRaskin: I have a reflow gun you can borrow if you want :p
<MichaelRaskin> I am not sure how much I will mess up if I try using it without supervision, that's a part of the problem…
<sphalerite> ah right, fair enough. I cannot provide competent supervision.
<eyJhb> MichaelRaskin: are you taking desktops with you in airplanes?
<MichaelRaskin> I mean, I have ever soldered things smaller than the standard breadboard connections
<MichaelRaskin> eyJhb: that's a BRIX
<eyJhb> Ahh, small cute machines
<MichaelRaskin> Yeah, except mine was not so «cute» for its time
<MichaelRaskin> i7-4770R
<MichaelRaskin> Its fan definitely harbours some regret it cannot be a helicopter blade
<eyJhb> If it was small, it was cute, if it was powerfull it had teeth :p
<MichaelRaskin> I mean, nowadays it could be probably regarded as just cute but noisy
<eyJhb> Sounds fun! Remember my first experience with server hardware + fans
<eyJhb> They try to take off! If they could they would
<MichaelRaskin> Although I guess for this size it might still be withing 2x of maximum consumer-achievable CPU power? «Small» ThreadRipper builds seem to need more space _just for the CPU itself_
<MichaelRaskin> But my current old ThinkPad W530 is comparable to BRIX, and has way more RAM than BRIX with one slot shot, so I do not use BRIX much nowadays
<eyJhb> The good ol' xx30 series
<eyJhb> Works like a charm
<MichaelRaskin> Frankly, I can name _one_ benefit over Asus
<MichaelRaskin> Battery (especially with second battery installed)
<eyJhb> And... Flexibility regards changing parts, availability of spare parts, ability to have cheap 3 year on-site repair, etc. :P
<MichaelRaskin> Well, I bought the ThinkPad used anyway
<MichaelRaskin> So this stuff about 3-year on-site repairs went by me.
<eyJhb> Same here, 2.5 years on-site repair :p
<eyJhb> So that shouldn't stop anything. The warranty follows the serial number, so I just upgraded to onsite for 400 DKK I think
<eyJhb> > DKK 400
<{^_^}> "1386004.000000 VND"
<eyJhb> > VND 1386004
<{^_^}> "52.668152 EUR"
<MichaelRaskin> I do not even know the age of this ThinkPad at the time of buying. And it was in Moscow, while I am mostly working in EU, which would probably complicate things anyway
<eyJhb> Ehhh... Lenovo doesn't care much as far as I know
<MichaelRaskin> As for flexibility — meh, CPUs are always soldered, and even in the thin Zenbooks changing RAM/storage is a breeze
elvishjerricco has quit [Ping timeout: 272 seconds]
elvishjerricco has joined #nixos-chat
pasukon has quit [Ping timeout: 272 seconds]
pasukon has joined #nixos-chat
<sphalerite> 5 days later… 5.46T resilvered, 99.28% done, 0 days 00:52:25 to go
<sphalerite> Got to love those SMR drives.
<MichaelRaskin> I don't think size-to-throughput problem is unique to SMR…
<ar> but this is taking it to the extreme
<ar> resilvering my 8×12T-drive raidz2 takes less time
<MichaelRaskin> _Every_ new generation of every storage technology takes this probleme to the extreme for this technology!
<MichaelRaskin> Yeah, SMR manages to be a new technology that makes things worse not better here, but not at a legendary level
<etu> sphalerite: oh, you run ZFS on SMR drives... :D
<MichaelRaskin> Doesn't help that large-storage has this ration on the edge of obviously-unusable, of course
* etu is still on 4T drives
<etu> Plan to upgrade any year now
<ar> i would love to upgrade to ssds
<ar> but 8×12T ssds would be prohibitively expensive
<sphalerite> etu: I didn't know about SMR when the hardware for this was being chosen
<sphalerite> etu: upgrade to more 4T drives :p
<etu> sphalerite: I plan to upgrade from mdadm/luks/ext4 to zfs while at it
<sphalerite> oooh yes
<ar> i plan to upgrade from zfs to 2-storage-node ceph cluster
<etu> This raid is at least 5-6 years old by now
<andi-> ar: two node ceph asks for troubles
<ar> andi-: why?
<ar> andi-: two storage nodes. i intend to run MONs on more nodes
<andi-> Ok
<andi-> That can work (and I did run one of those for many years) but being able to reboot a node without loosing write ability is even better :)
<eyJhb> MichaelRaskin: Once had to clean a Asus laptop. The battery connector came off the motherboard because the quality was so bad
<eyJhb> NEVER tried that, and I think I am up in the hundres of laptops taken apart.e..
<MichaelRaskin> Was it an X and how old?
<ar> MichaelRaskin: not sure how it is right now, but at least back around 2011-2013 thinkpad warranty was worldwide
<eyJhb> MichaelRaskin: I am actually quite sure it was a Asus Zenbook, not sure of the age
<ar> MichaelRaskin: as in, i had a thinkpad T420s imported from US, and polish lenovo warranty services didn't give me any trouble and even honoured the on-site next-business-day
<eyJhb> ar: and the xx30 series is from 2012! :D
<eyJhb> Yeah, it is awesome. It is one of the key reasons I want another Thinkpad with on-site
<eyJhb> ALways nice to get stuff replaced if it breaks too easily
<MichaelRaskin> Hmm, that might mean I bought that Thinkpad already past its warranty
<eyJhb> When did you aquire it?
<MichaelRaskin> I am not sure, it might have been 2016
<eyJhb> Ahh. then that would make sense...
<MichaelRaskin> Ah wait, I have mtn log
<MichaelRaskin> Yep, late 2016
<eyJhb> I have saved my invoice as x240
<eyJhb> WEird
<eyJhb> Bought mine in mid 2014
<eyJhb> Holy hell, it is actually 6 years old :| It had birthday yesterday or something
<MichaelRaskin> The thing I started to appreciate about Asus after getting a Thinkpad (hm, might have been first non-Asus I used closely in ten years?) is that Asus cuts corner on materials compared to Thinkpads and kind of proportionally to price difference, but Lenovo cuts corners on sanity
<eyJhb> Sanity?
<eyJhb> ALso, the materials for my x230 sucks
<MichaelRaskin> After using UEFI setup on Asus and UEFI setup on Lenovo, I would be happy if like Asus became a monopoly on writing UEFI implementations.
<MichaelRaskin> Also, it is just beyond absurd to have Intel GPU drive dock video outputs and Nvidia drive the left-hand-side video connectors
<ar> the only issue i have with lenovo/thinkpad uefi, is that adding boot entries directly from it isn't a thing
<MichaelRaskin> Which is like the only thing there is to configure about UEFI!
<eyJhb> Directly from it?
<ar> yup
<eyJhb> PRetty sure I installed NixOS onto my x230, without issues
<ar> it was a thing on the only dell laptop i had - e7440 - but i hated that one for a lot of other reasons
<eyJhb> What is "it" ? BIOS UEFI?
<MichaelRaskin> Asus allows you to configure the partitions, and paths relative to the partitions, just as it should be
buckley310 has quit [Quit: Connection closed for inactivity]
<MichaelRaskin> Has anyone checked whether GitHub's gh is good enough to dump a backup of all discussions in NixOS GH org easily?
<Mic92> MichaelRaskin: for pull requests/reviews I use the graphql api to reconstruct all information from the web-interface.
<Mic92> in nix-review
<eyJhb> Any matplotlib experts? - https://i.imgur.com/lYzPd6B.png doesn't show the entire figure, only a portion of it
bqv has joined #nixos-chat
<MichaelRaskin> Hm, no idea, but shouldn't you also specify the figure ranges?
<eyJhb> MichaelRaskin: no clue, I do not use matplotlib that often and am somewhat frustrated that there seems to be no easy way to get the axes "inside" the actual figue etc.
<MichaelRaskin> When I have detailed ideas what it should look like, I just use Asymptote
<eyJhb> Yeah, but I can't do that atm. :/ And also, I need to do "more" so using Asymptote would require me to either 1. export data from python and import/parse it in asymptote or 2. Implement probability functions in asymptote
<MichaelRaskin> 2 is not that bad of an idea but not if you already have all computations working in Python, of course
<MichaelRaskin> Asymptote, apparently, has csv support built-in (and of course array literals are also cheap to generate)
<eyJhb> MichaelRaskin: And if I have unlimited time, then I would do that. But I have atm. 8-10 hours a day of uni work... So it needs to be "fast" as well
<eyJhb> But I got it to work now `fig, ax = plt.subplots(); ax.set(xlim=(-2,2), ylim=(-2,2)); ax.axes.set_aspect('equal')`
<eyJhb> Semicolon to one-line it :p
<MichaelRaskin> This is in line with what I expected (re: setting ranges), and does it still draw axes through (0,0)?
<eyJhb> Currently at this point now - https://i.imgur.com/ZtPpouR.png
<eyJhb> But it is a hell of a lot of commands, to get it to draw axis in the center, instead of to the side
<MichaelRaskin> Ah
<eyJhb> https://i.imgur.com/1O3HOZF.png just feels off
<eyJhb> Might just be me
<joepie91> eyJhb: it runs outside the square on the left
<joepie91> and top
<MichaelRaskin> If you need a lot of graphs, uiformity matters, details whatever
<joepie91> definitely does not look centered
<eyJhb> joepie91: yeah, I can see that. Doesn't really make sense :p
<eyJhb> Guessing it is just a matter of the draw
<eyJhb> MichaelRaskin: Agreed
<MichaelRaskin> If you need to do 8 hours of practice computations, give up on aesthetics as long as readability is fine
<eyJhb> joepie91: It looks better in the center axis image, but I am sure it is just some minor thing in matplotlib. Because it "cannot" go out of the box.
<eyJhb> MichaelRaskin: yeah, that is what I am going for. Trying to make these exercises digital, as I hope I can get better at Python and then do othe exam faster
<eyJhb> do the*
<eyJhb> Because, by hand I will never get done
<eyJhb> My back all of the sudden started to kill me, yay
<MichaelRaskin> Yeah, surely carefully tuning the image in Asymptote (or even with matplotlib options) makes sense when the best-case amount of person-seconds looking at it is above 60.
<MichaelRaskin> When it is just one glance by you «seems right» and one glance by the grader «and calculate that distribution they did», defaults are fine. Also, defaults might make it more uniform with what _other_ submissions contain, I guess?
<eyJhb> I have just defined a pltDefaults(ax): so that I can turn it on/off for all figures as needed
<MichaelRaskin> Now I want to invent some absurd reason why calling something Peripheral limit theory as a dual to Central limit theorem could make sense. Maybe Kolmogorov's 0-or-1 law?
<MichaelRaskin> (prompted by PLT being similar to CLT)
NinjaTrappeur has quit [Quit: WeeChat 2.9]
NinjaTrappeur has joined #nixos-chat
ixxie has joined #nixos-chat
<gchristensen> are there any nice tools for logging every time a specific program is exec'd, and by what?
<MichaelRaskin> Hmmm. How much control you have, can you just replace it with a logging script checking own PPID?
<gchristensen> yeah that might be the thing. I have total control
<MichaelRaskin> I assume you do not want to mess with ptrace-ing the entire system
<gchristensen> noooot really.
<MichaelRaskin> I guess there are some inotify-based options but inotifywait does not print the process information
<joepie91> there's auditd
<joepie91> if you can find the docs
<V> I was going to suggest audit
<V> that has exactly that support
<V> more generally it can log when specific files are accessed, etc
<gchristensen> nice
<joepie91> yeah auditd is like, almost verbatim what you want... it's just that every time I've tried to use it, I was unable to find half the docs, so it's a bit of trial-and-error to set up
<joepie91> I've used it in the past mainly to debug weird network issues
<MichaelRaskin> Hmm, is auditd just as costly as ptrace, though?
<joepie91> It Has Been Claimed(tm) that it can run relatively unobtrusively
waleee-cl has joined #nixos-chat
<gchristensen> hehe, security.audit.enable Type: one of <bool>, <bool>, "lock"
<infinisil> Oh that's not very good
<infinisil> gchristensen: https://github.com/NixOS/nixpkgs/pull/99471 :)
<{^_^}> #99471 (by Infinisil, 9 seconds ago, open): lib/types: Fix type description of bool enum values
<gchristensen> oh!
<gchristensen> hmm audit's logs show the ppid but by the time I can look at it, that ppid is gone
<MichaelRaskin> If the parent actually waits for reply from the process, you know a wrapper script has all the time it needs to inspect parent
<sphalerite> gchristensen: I've used execsnoop, from bcc, occasionally.
<gchristensen> that might be a little tricky. I'm really unsure where the program (/bin/sh) is being called from. I think I'll give a wrapper a shot :)
<sphalerite> oh yeah, same ppid checking issue I guess
<MichaelRaskin> I wonder if we ever reach some good solution for system() (3) problem
<gchristensen> eliminating the PATH variable and embedding Nix closures in to the libc's exec path finder
<MichaelRaskin> For the record, I do not consider adding all the possible tools I might want to spawn from Vim into Vim's closure a good solution
<MichaelRaskin> And the problem with system() is not that, but circular depedency between some sh and some libc
<gchristensen> okay fine, little a inheritance, as a treat
<MichaelRaskin> I guess it doesn't actually matter for my case, as I can just override what is used for /bin/sh per jail, which means almost the same as per application in my setup…
<MichaelRaskin> But it sometimes breaks Hydra etc.
<gchristensen> abathur: I wonder if resholve should touch the shebang too
<abathur> why? would it just be saving a patchshebangs step, or some other purpose?
<gchristensen> good point, I can just add a patch shebangs to my pkgs.resholve
maxdevjs has joined #nixos-chat
<abathur> fair, though; it probably makes sense for the nix resholve function to have an option to patch or not
<abathur> https://github.com/mabam/CAP/blob/ab979f639dcd9b8901122e197ec8fdfae2362db2/contrib/MacPS/macps-22.shar#L2-L18 :] I'm not quite sure if I didn't know shell archives were a thing, or if I've seen it before and just forgot, but this lit up my novelty receptors either way
betawaffle has quit [Ping timeout: 272 seconds]
sorear has quit [Ping timeout: 272 seconds]
manveru has quit [Ping timeout: 260 seconds]
liszt has quit [Ping timeout: 272 seconds]
kalbasit has quit [Ping timeout: 272 seconds]
claudiii has quit [Ping timeout: 272 seconds]
manveru has joined #nixos-chat
claudiii has joined #nixos-chat
liszt has joined #nixos-chat
kalbasit has joined #nixos-chat
betawaffle has joined #nixos-chat
sorear has joined #nixos-chat
<sphalerite> abathur: ha, haven't seen that style before. I've only seen them as shell scripts with a tar or similar appended.
<gchristensen> abathur: hmmm is there a way to tell resholved this is okay actually? https://gist.github.com/grahamc/bcdaf76e72188b07760e4d632c16a12d
<abathur> yeah; aside from adding it to the inputs you can allow resholved_inputs:<osh path>, via the command-line or Nix api
maxdevjs has quit [Remote host closed the connection]
maxdevjs has joined #nixos-chat
<gchristensen> hmm
<gchristensen> let me push my WIP and show you what I'm shootingfor
<abathur> I've been picking at an API rework that'll hopefully make more sense, too; the initial names were a little hasty to avoid overthinking until I had a better sense of what options would turn up and what sort of patterns/groups would emerge
<gchristensen> abathur: https://github.com/grahamc/nixos-config/commit/e13da65cb6474a3084f48855d4c57f3a5a7c5cdb I suppose maybe I could add an attribute to the "resholve" function to allow "target"
<abathur> I think if you add oil to your inputs in snoop/default.nix?
<abathur> regarding the "allow" approach, there's just a delta between your nix resholve func and the one in the resholve repo (https://github.com/abathur/resholve/blob/master/resholve-package.nix, example at https://github.com/abathur/resholve/blob/master/ci.nix#L14-L26
<gchristensen> ah!
<abathur> guess that worked?
<gchristensen> taking a look at it now :)
<abathur> may not make sense for you to adopt that Nix API in the short run if it'll cause you much churn, since I'm midstream on a revision
<gchristensen> nah, it is good for now :)
<gchristensen> okay, rebooting to see what kind of havoc this causes.
<abathur> <3 gchristensen
<{^_^}> gchristensen's karma got increased to 343
<abathur> fair chance to ask; tentatively splitting the hodgepodge "allow" into "fake" (for pretending some things are defined), "fix" (telling resholve to fix something it wouldn't b
<abathur> gah
<gchristensen> huh, /bin/sh being oil doesn't break firefox for some reason this time
<gchristensen> also this snoop script has found bugs in nixos
<abathur> ooo
cole-h has joined #nixos-chat
<abathur> fair chance to ask; tentatively splitting hodgepodge "allow" into "fake" (pretend some things are defined), "fix" (tell resholve to fix something it wouldn't by default), and "keep" (~ignore things it would warn about); so `--allow "unresholved_inputs:<osh_path>"` tentatively -> `--keep "<osh_path>"
<gchristensen> hmm I that seems okay I think
<gchristensen> #99492
<{^_^}> https://github.com/NixOS/nixpkgs/issues/99492 (by grahamc, 7 seconds ago, open): Powertop refers to /usr/bin/hcitool
<abathur> nice
<cole-h> warning: unknown setting 'experimental-features'
<cole-h> :P
<abathur> gchristensen: were you intentionally looking for things like this, or was it an incidental discovery?
<gchristensen> I was trying to figure out why using osh as /bin/sh caused firefox to stop resolving domains. it turned out to have just been a fluke that particular boot, that networking wasn't working right
<sphalerite> gchristensen: hm, was resolution failing immediately or just hanging?
<gchristensen> I think it was immediate
<sphalerite> oh ok, then it's not the thing I've seen a couple of times.
<abathur> per your request on ~traced execution, it may help to keep a list of known cases of unresolved and/or hardcoded external deps as we find them (whether in binaries or scripts) for testing? a lot of the paths I've looked down seem like dead-ends from a cross-platform perspective, but I have spotted at least two approaches, and you've added one here
rajivr has quit [Quit: Connection closed for inactivity]
__monty__ has joined #nixos-chat
endformationage has joined #nixos-chat
<ldlework> eyJhb: While emacs is infinitely extendable and you can spend a lot of time on it if you want, I feel like those two were selling you something else.
cjpbirkbeck has joined #nixos-chat
cole-h has quit [Ping timeout: 240 seconds]
<sphalerite> joepie91: re your tweet from yesterday: but muh reeeeeent
<joepie91> sphalerite: which of them? :P
<sphalerite> joepie91: about software companies' core business
<joepie91> ahhhh
<joepie91> yeah...
<__monty__> TLDR?
<sphalerite> __monty__: did you really just ask for a tldr of a tweet?
<joepie91> lol
<samueldr> that's not proper use of twitter here
<__monty__> I don't like visiting twitter.
<drakonis> nixpkgs is so massive that github is having trouble with processing data for the insights tab
<samueldr> you have to go through 20+ tweets goint 1/ 2/ 3/
<drakonis> i cant visualize the commit growth throughout its history with it anymore
<drakonis> which repo has more activity than nixpkgs right now?
<drakonis> okay it finished processing the commits, its still trending up
<sphalerite> joepie91: do you mirror your tweets on a mastodon/whateverelsefreefederatedthing instance?
<sphalerite> joepie91: also, just saw the encoding grid donate thingy on http://cryto.net/donate/. I'd be happy to donate some spare cycles on my build server if that ever gets going :)
<infinisil> Bad idea: Once flakes can be used in nixpkgs, we could have a separate nixpkgs-sources repository that contains all package sources. That one could be much faster moving than nixpkgs itself, and much more automated
<infinisil> nixpkgs would depend on nixpkgs-sources, which would be updated like every day
<sphalerite> infinisil: all versions of all packages, right? :D
<infinisil> sphalerite: Maybe even multiple versions yeah!
<infinisil> And one latest link
<joepie91> sphalerite: I don't currently mirror tweets anywhere, no - don't want another piece of fragile infrastructure to maintain :P
<joepie91> sphalerite: and heh, that encoding grid will come to exist Some Day, but probably quite far into the future
<joepie91> thanks though :D
<joepie91> (the site is horribly outdated)
<sphalerite> joepie91: well, you know, you can use somebody else's instance :p
<joepie91> sphalerite: I'm the one people will bother when it stops mirroring :P
<drakonis> infinisil: its not a bad idea!
<drakonis> but then, i'm generally quite insane myself
<infinisil> The bad thing about it I had in mind was that you now have a package definition split into different repos
<infinisil> So you can't just PR a new package to nixpkgs with a single reproducible file
<infinisil> But this idea also has a bunch of pretty big advantages
<infinisil> easy version updates, much less nixpkgs PR traffic, more automatable updates, multiple versions
<infinisil> Easy version overriding
<drakonis> hmm
<drakonis> you really would not
<drakonis> i have an approach that would be actually quite interesting
<drakonis> you'd have each release flake define which version of the packages they need
<drakonis> it would make it vastly easier to propagate security fixes
<f0x> hmm morph says it's for "managing *existing* NixOS hosts", but would it be possible to deploy to a machine that's still booted into the installation ISO?
<drakonis> if there's any changes specific to that default environment, it'd go on the repo
<f0x> so boot ISO, format disks, push morph, reboot into fresh server
<infinisil> f0x: I don't know how morph works exactly, but I wouldn't expect that to work unless it does something special to handle such a case
<infinisil> drakonis: What's a "release flake"? A nixpkgs release one?
<drakonis> yes
<aanderse> is anyone running 20.09 with kde? it seems like a bunch of kde apps are broke
<aanderse> konversation, krename, more
<drakonis> ie: you have nixos-20.09 defining which packages are the default versions
<drakonis> in a world where you have pretty much every version on the same repo
<drakonis> pin the versions each release has
<drakonis> let's say a cve comes up, when you fix it, currently you'd have to propagate it to every release that still has the affected software
<drakonis> this method would get rid of that issue as anyone who's pinning that specific software version would get the updates
<drakonis> simply update your current release package set to get the updates
<drakonis> its not really a big difference from what's being done today
<drakonis> from how its done right now
<joepie91> f0x: that wouldn't work because when you're in the live ISO, / points to the in-memory FS
<joepie91> or well, it'd "work" in the sense that it'd deploy your config to the in-memory system :P
<joepie91> I typically install it with a dummy config
<joepie91> (make sure to update the stateVersion)
<joepie91> and then once booted into that, I deploy with morph
<f0x> joepie91: ah, right. Wondered if perhaps it could deploy to /mnt instead
<joepie91> not that I know of
<joepie91> but you need to SSH into it anyway, and detach the ISO at some point
<joepie91> so I'm not sure that that would actually be any faster than just installing the dummy config and rebooting :P
__monty__ has quit [Quit: leaving]
<ldlework> drakonis: I got my Ammo Switcher module working
<ldlework> wanna try it?
<samueldr> I'm annoyed at how networking can be so... weird
<samueldr> like right now the existing connection to IRC works
<samueldr> but somehow opening new connections broke?
<ldlework> samueldr: when dns goes out but not your connection
<samueldr> shouldn't be DNS
<samueldr> unless the browser literally does a DNS lookup every link I follow
<samueldr> whatever happened now resolved itself
<colemickens> did it systemd-resolved itself?
<drakonis> its very nice
<samueldr> I don't know if systemd was involved
<samueldr> it might have been an external networky thing
<samueldr> I really hate networking
<samueldr> mostly because I don't grok it
<infinisil> I took a bunch of network courses
<infinisil> I still don't grok it
<samueldr> now most of what I know is from research when things don't work or I try to get things working
<samueldr> one of the reasons my network is served with openwrt routers still
<samueldr> the work needed to setup a router using NixOS would be much greater because I'd also need to learn more about networking
drakonis has quit [Quit: ZNC 1.8.1 - https://znc.in]
ixxie has quit [Remote host closed the connection]
<Ashy> nscd has caused problems for me a fair bit on nixos
<Ashy> i had it disabled for a while but that breaks something else now so i just manually stop and start the systemd service for it when i notice dns being weird now
<samueldr> I'm pretty sure it wasn't dns
abathur has quit [Quit: abathur]
abathur has joined #nixos-chat