<gchristensen>
that is a footgun: I made a typo in setting up a new box with nixops, setting it to deploy to foo.bar.cmo, which resolved to ... my server ... which nixops dutifully ssh'd in to, copied the new config, and did a rebuild-switch :P
<coconnor>
ouch
<cransom>
ahaha. oooops indeed.
<gchristensen>
good news: out of band SOL console access + nixos-rebuild switch --rollback, and it was fixed in a couple minutes :)
<infinisil>
gchristensen: Why would you set up a DNS to route all domains to your server??
<Meizikyn>
so the current implementation of aufs for docker is docker specific? (I don't know shit about docker)
<gchristensen>
infinisil: I just ... I dunno!
<Meizikyn>
<embarks on quest to package>
<{^_^}>
[nixpkgs] @Ericson2314 merged pull request #37395 → lib: Factor in tiny bit of `meta.platform` checking → https://git.io/vxGMz
<{^_^}>
→ 26fac59c by @coreyoconnor: gapi-ocaml: 0.3.4 -> 0.3.6. Also change to use opam/jbuilder for build.
<{^_^}>
→ a140fe83 by @coreyoconnor: google-drive-ocamlfuse: 0.6.21 -> 0.6.25.
<{^_^}>
→ f71a636d by @Mic92: Merge pull request #37385 from coreyoconnor/google-drive-ocamlfuse-update
generic has joined #nixos
nikivi has joined #nixos
<infinisil>
ottidmes: I clearly don't understand the usecase here, can you tell me?
<infinisil>
ANd gchristensen for that matter
<ottidmes>
infinisil: Well, in my case everything with the same domain is destined for the same server, and I have nginx handle the subdomains
<infinisil>
But gchristensen said "all domains" so i assumed this also includes stuff like google.com, but maybe i understood wrong
<ottidmes>
infinisil: To clarify, everything that goes to example.com, goes to a certain IP, including all its subdomains, the semantics of what happens with the subdomain is up to in my case nginx
<ottidmes>
infinisil: Yeah, I would agree that that would be bad :P
pkill9 has quit [Ping timeout: 255 seconds]
<infinisil>
ottidmes: Well gchristensen said his typo was foo.bar.cmo, which would indicate that he routes every invalid domain to his own server!
<ottidmes>
infinisil: That true, yep, in that, but why? :P
<ottidmes>
infinisil: I only have such a wildcard for *.localhost, but that points, as you would assume, to 127.0.0.1
<gchristensen>
because I'm a silly goose
<infinisil>
But like, why did you do it in the first place?
_null_ has joined #nixos
<infinisil>
What was the intended usecase?
<gchristensen>
I honestly don't know
<gchristensen>
it may just be a mistake
<infinisil>
Haha
<gchristensen>
may have just started as a mistake*
Meizikyn has quit [Ping timeout: 260 seconds]
freeman42x]NixOS has quit [Ping timeout: 252 seconds]
muzzy has joined #nixos
<gchristensen>
w00t I have extremely hacky nixops <-> Packet spot instance support going
sahabi_ has joined #nixos
jrolfs has joined #nixos
sahabi has quit [Ping timeout: 240 seconds]
hakujin1 has quit [Ping timeout: 240 seconds]
mounty has quit [Ping timeout: 240 seconds]
vidbina has joined #nixos
<{^_^}>
[nixpkgs] @dezgeg pushed to master « armTrustedFirmware: Remove targetPlatforms like was done in U-Boot »: https://git.io/vxG7J
hakujin1 has joined #nixos
paraseba has quit [Ping timeout: 252 seconds]
<{^_^}>
[nixpkgs] @dezgeg pushed 2 commits to release-18.03: https://git.io/vxG7s
<{^_^}>
→ 607fc823 by @dezgeg: nss: Fix aarch64 build
<{^_^}>
→ 5abaf2a9 by @YorikSar: chromium: 65.0.3325.146 -> 65.0.3325.162
<{^_^}>
[nixpkgs] @dezgeg pushed commit from @thefloweringash to release-18.03 « Chromium: fix skia build on aarch64 »: https://git.io/vxG7l
<{^_^}>
[nixpkgs] @dezgeg pushed commit from @YorikSar to release-17.09 « chromium: 65.0.3325.146 -> 65.0.3325.162 »: https://git.io/vxG7u
maximiliantagher has quit []
sahabi_ has quit [Quit: Leaving]
paraseba has joined #nixos
maximiliantagher has joined #nixos
<ottidmes>
Does anyone perhaps know how you disable Firefox from logging stuff in the journal? Quite annoying to see it spammed with JavaScript errors
mounty has joined #nixos
hakujin1 has quit [Ping timeout: 276 seconds]
abathur has quit [Quit: abathur]
hakujin1 has joined #nixos
<coconnor>
ottidmes: do you launch firefox in xsession by chance?
<clever>
i believe it would inherit the stdio of the window manager, and wind up in display-manager.service's logs anyways
<clever>
the only way to not do that, is to run it from a terminal
<clever>
or wrap it with a bash script that redirs to null
hakujin1 has quit [Ping timeout: 252 seconds]
<ottidmes>
clever: I launch it via rofi, but the logs say xsession, so I guess so
hakujin1 has joined #nixos
<ottidmes>
clever: ^
<ottidmes>
Ah come on..
<ottidmes>
coconnor: ^
<clever>
ottidmes: check `ps -eH x` while firefox is running, and follow the tree up
<ottidmes>
clever: Would simply wrapping it and writing stdout and stderr to /dev/null work? I could try, but I thought that that did not work
<ottidmes>
clever: I see /nix/store/wj4zc68d84202l64y9nisklmnqsf4pky-firefox-unwrapped-58.0.2/bin/.firefox-wrapped and some contentproc children, the parent is systemd
hakujin1 has quit [Ping timeout: 256 seconds]
<clever>
ottidmes: what is the parent of firefox-wrapped?
<ottidmes>
clever: systemd, at least according to that ouput of ps
<vaibhavsagar>
it's blocking my IHaskell upgrades so I'm sticking with an old revision of nixpkgs for now :)
<foldingcookie>
"upgrading 'font-adobe-75dpi-1.0.3' to 'font-adobe-100dpi-1.0.3'" how would one explain to nix that this isn't a valid upgrade path
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #37398 → lib: Make platform predicates greppable → https://git.io/vxGN7
shabius has quit [Remote host closed the connection]
<zie[m]>
when I nix run nixpkgs.pkgconfig , I can't actually run pkg-config.. I just get an error saying install it.. nix-info and output: https://pastebin.com/kcjUQ68n
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #37399 → lib: Messed up `or` operator precedence → https://git.io/vxGAW
<{^_^}>
[nixpkgs] @andir pushed 5 commits to release-17.09: https://git.io/vxZfC
<{^_^}>
→ 30e84fac by @andir: rust121bin: 1.21.0 -> 1.24.0
<{^_^}>
→ 3b34504f by @taku0: firefox-bin: 59.0 -> 59.0.1
<{^_^}>
→ a8c8d062 by @taku0: firefox: 58.0.2 -> 59.0.1
<fearlessKim[m]>
I get a similar error than yesterday but this time I don't have a core debug in my nixops folder. `server> copying path '/nix/store/9lharz6d9i2zp92zl6w4v7ifks15m775-qemu-2.11.1' to 'ssh://root@192.168.122.222'...
<fearlessKim[m]>
client> warning: dumping very large path (> 256 MiB); this may run out of memory
<fearlessKim[m]>
client> error: out of memory`. It happens only when I deploy with `environment.systemPackages = [ pkgs.qemu ];` (export GC_MAXIMUM_HEAP_SIZE=10G didn't help either).
<fearlessKim[m]>
(trying to add qemu-agent to libvirtd nixops images)
<Myrl-saki>
clacke[m]: Okay, background. Basically, I want to use `runhaskell ./Setup` rather than compiling Setup.hs
<Myrl-saki>
clacke[m]: This causes a rebuild on GHC, which takes hours. I'm trying to avoid that rebuild.
<Myrl-saki>
Basically, "use `runhaskell ./Setup` on packages, but not on GHC."
<Myrl-saki>
In `generic-builder.nix`, that is `setupCommand = "runhaskell ./Setup";`
<clacke[m]>
so then you would like to override setupCommand on each of the haskell packages you want to build?
<Myrl-saki>
clacke[m]: On every possible haskell packages.
<Myrl-saki>
Except for ghcjs.
<Myrl-saki>
Err
<Myrl-saki>
Except for ghc, and, well, ghcjs.
<Myrl-saki>
Basically, except for the compilers.
<clacke[m]>
and you can do that by overriding it in some central place, but that would affect haskell itself, so you want to pick that from the unmodified pkgs
<Myrl-saki>
clacke[m]: What do you mean?
<clacke[m]>
I'm just trying to understand what you are saying and what you are doing.
<clacke[m]>
So I'm assuming you found some central place that you can poke so that all packages are affected.
<Myrl-saki>
clacke[m]: Yep, which is the generic builder.
hamishmack has quit [Quit: hamishmack]
<clacke[m]>
But then you want to put back the original unpoked things for ghc and ghcjs.
<Myrl-saki>
I wonder if I can do some kind of introspection?
<Myrl-saki>
"What package am I building" kind of thing.
<clacke[m]>
Well sure, in your alternative setup you could do `setupCommand = if (builtins.elem name [ "packages" "I" "want" "unpatched"] ) ''original setup exactly'' else ''runhaskell ./Setup'';`
<srhb>
How do I get logs for a build I just did with Nix 2.0? It appears to attempt to download them, and they're not available (of course, since I modified the package.)
<Myrl-saki>
Back.
<Myrl-saki>
clacke[m]: srhb: I'm guessing ghcjs indirectly relies on a haskell package.
Lears has joined #nixos
rauno has joined #nixos
<Myrl-saki>
s/ghcjs/ghc/
<fearlessKim[m]>
I don't get the "out of memory file" ? isn't is possible to copy files in chunks to stay wuthin memory bounds ?
<Myrl-saki>
I think it's hscolour.
<Myrl-saki>
clacke[m]: srhb: I was right. :)
<srhb>
woo :)
<Myrl-saki>
srhb: clacke[m]: setupCommand = if builtins.match "(ghc|hscolour).*" pname != null then "./Setup" else "runhaskell ./Setup";
<Myrl-saki>
Hopefully that will improve build times.
<tA63>
I'm trying to run a script that how references to /bin/rm and other programs, whats the correct way to go about changing this to run in NixOS?
<Myrl-saki>
gdi, I'm pretty sure I know what happened.
<srhb>
tA63: Substitute them all with actual nix store paths
<Myrl-saki>
Yep, I'm pretty sure I know now.
<tA63>
If I was building a pkg, how do I go about getting the store path on a system?
pie_ has joined #nixos
<Myrl-saki>
srhb: clacke[m]: I have a feeling I know why Setup.hs is compiled now.
<srhb>
tA63: Usually it's all just strings. So eg. "${coreutils}/bin/rm" is the store path for rm
<tA63>
Ahh cool, thanks
<srhb>
tA63: Yeah, it would be pain and horror if packages didn't toString to their store paths like that :-)
<Myrl-saki>
srhb: clacke[m]: runhaskell will use the Cabal version that your ghc is built with, I think.
<tA63>
Slowly learning things haha, still gotta get home-manager set up nicely
MercurialAlchemi has quit [Ping timeout: 268 seconds]
MercurialAlchemi has joined #nixos
ashgillman has joined #nixos
<srhb>
tA63: For instance, I have this in my home-manager: systemd.user.services.setxkbmap.Service.ExecStart = "${pkgs.xorg.xkbcomp}/bin/xkbcomp -merge $HOME/.xkb/base.xkb $DISPLAY";
<srhb>
tA63: (Which will automatically make xorg.xkbcomp a dependency of that systemd unit)
rauno has joined #nixos
<srhb>
I should nixify base.xkb too of course.
<tA63>
Ahh neat
<tA63>
Ok so I think I'm doing something dumb. But if I have a shell script to install something, first thing I want to do is replace all he hardlinked things with where they are in nix right? I still havent found the path to my rm in actuality, I tried to evaluate ${coreutils}/bin/rm and it was blank
<Myrl-saki>
What other uses do you guys think Nix can have?
<{^_^}>
→ d6d140f1 by @joelthompson: hologram: 2017-01-30 -> 2018-03-19
<{^_^}>
→ fe2e4d6f by @joelthompson: hologram: Enable configuring LDAP authorization
rauno has quit [Remote host closed the connection]
<fearlessKim[m]>
any reason why multiple swap definitions can't be merged together ? `The unique option `swapDevices.[definition 1-entry 1].device' is defined multiple times, in `/home/teto/dotfiles/nixpkgs/configuration.nix' and `/home/teto/nixpkgs/nixos/modules/config/swap.nix'`
<srhb>
Myrl-saki: But if you check out the nixcon 2017 talks, there's some on the subject :)
<srhb>
Myrl-saki: Also the tweag.io Nix/Bazel Haskell stuff looks really interesting, but I didn't dive in much yet.
mudri has joined #nixos
<tA63>
Sorry for all the newbie questions, but whenever I get a script to run, I end up with a line saying a certain file doesnt exist, but it definitely exists at that path
<joepie91>
tA63: my general advice is that you should expect to spend a few weeks upfront learning how everything works in NixOS (/ Nix), essentially relearning Linux... but once you have, it's worth it :P
<joepie91>
it can be pretty frustrating at first because nothing works quite how you expect it to
<tA63>
Yeah I figured that'd be how it goes
dvim has joined #nixos
<joepie91>
and like, there's good reasons /why/ it doesn't work how you expect it to, but the mismatch is still there :P
<etu[m]>
tA63: It's perfectly normal to want to back out of Nix, I wanted to but then I learned things and started contributing instead :D
winem_ has joined #nixos
roberth has joined #nixos
civodul has joined #nixos
winem_ has quit [Remote host closed the connection]
<srhb>
Anyone up for some psychic debugging? I have a derivation (which I cannot share) that, when I build it using nix-shell and genericBuild, setting out to somewhere in tmp, has certain files in its /tmp/foo/sbin folder. When I build it using nix-build, the store path does not contain these files. Any pointers?
<srhb>
manveru: I am. I guess I could try turning that off and see if it makes a difference.
<srhb>
I suppose the shell might also not be sufficiently pure despite --pure, subtly altering the build.
<manveru>
yeah... it's not pure since it still loads our rc files
<srhb>
The build has a 20 minutes turnaround time, so I might just need something stronger than coffee...
zzamboni has quit [Ping timeout: 255 seconds]
asuryawanshi has joined #nixos
zzamboni has joined #nixos
NinjaTrappeur has quit [Quit: WeeChat 2.0.1]
zzamboni has quit [Client Quit]
zzamboni has joined #nixos
raynold has quit [Quit: Connection closed for inactivity]
<fearlessKim[m]>
trying to create a swap on /dev/fucking_swap`fallocate: fallocate failed: No space left on device`, it's because /dev is full:`devtmpfs 396M 396M 0 100% /dev` . Does nix do sthg to set the maximum size of /dev ?
<clacke[m]>
tA63: My advice is not to run smack into NixOS but start with Nix on a system that works for you and that you understand.
zzamboni has quit [Client Quit]
<clacke[m]>
Making your computer work while figuring out Nix is a bit much to ask.
<TimePath>
+1 to that, I migrated to nixos by setting up a VM how I like it before committing
spear2 has joined #nixos
<katona>
Hi, why does hydra trunk have so many failing jobs for quite some days now?
<joepie91>
katona: it recently had its hardware upgraded, not sure if that may have a shared cause
zzamboni has joined #nixos
<joepie91>
(yesterday, iirc)
<katona>
joepie91: Oh I see. (That's why there was a server error :O)
paraseba has quit [Ping timeout: 240 seconds]
asuryawanshi has quit [Remote host closed the connection]
asuryawanshi has joined #nixos
<katona>
But I think the 12000+ failed jobs have been going on for at least five days I think and I can't NixOS rebuild into it with a pretty minimal config even though I really liked testing bleeding edge apps :/
<{^_^}>
→ 366c79e1 by @teto: pythonPackages.neovim: 0.2.3 -> 0.2.4
<{^_^}>
→ b41c299e by @dotlambda: Merge pull request #37400 from teto/neovim_python
<etu[m]>
srhb: That I know... I just hope that's more unlikely...
<etu[m]>
But since it's possible it will happen.
<srhb>
Indeed. :P
<joepie91>
foldingcookie: srhb: etu[m]: obligatory warning that IPFS does _not_ magically persist data, it is a distribution mechanism, not a storage mechanism
<joepie91>
this doesn't mean that it can't be useful, but too many people have the expectation that you drop data on it and it'll magically keep existing
<joepie91>
yes, as far as I'm aware Filecoin is just as broken as every other proof-of-storage system
<srhb>
joepie91: It solves nothing in either case. The issue is that upstream isn't actually promising not to rewrite history :-P
<infinisil>
joepie91: It doesn't even exist yet? How can it be already broken? And how are others broken?
<joepie91>
infinisil: because the theory underlying it can't work, or at least I've not seen anybody plausibly prove that it can
<joepie91>
infinisil: the problem is that nobody has figured out how to reliably prove third-party storage without keeping a copy yourself
<etu[m]>
Blockchain for sourcecode-releases :p
<joepie91>
infinisil: lots of people *claim* that they have, but have then overlooked certain threat models, either intentionally or accidentally
<joepie91>
it's a super hard problem to solve
<{^_^}>
[nixpkgs] @asppsa opened pull request #37427 → emscripten/binaryen: make binaryen track the current emscripten version → https://git.io/vxZng
<joepie91>
therefore anything that's based on the premise of "prove that a third party stored this data", cannot work
<infinisil>
joepie91: I mean they won't be able to prove 100% that your data will always be available (as long as you pay), but I imagine it's only a percentage, e.g. you pay about $0.001/hour for 99% availability, $0.005/hour for 99.9%, etc.
<infinisil>
That's how i think it's gonna work
<joepie91>
infinisil: but you can't prove that either.
<infinisil>
The more you pay the more nodes will store your data
<joepie91>
like, it's not a percentage thing
<joepie91>
it's an "everybody's model fails in these particular attack scenarios that can be executed an arbitrary amount of times" thing
<infinisil>
Yeah it's all probabilistic of course, the network should just distribute it as good as possible
<joepie91>
hence, it's an X% chance of 100% data loss, not a 100% chance of X% data loss
<joepie91>
which is what makes it functionally useless
<etu[m]>
But if a client say it store something but actually doesn't, nobody can prove that that client actually don't store it until you try to fetch that data from that client.
<srhb>
Proposed tagline: Store your data maybe for cheap, probably.
<infinisil>
joepie91: Why would you lose all data?
<joepie91>
srhb: sounds about right :)
<etu[m]>
:D
<joepie91>
infinisil: because you have absolutely no way to verify that *anybody* is storing your data, reliably.
<infinisil>
joepie91: you don't? You can just ask ipfs for the nodes that store a certain hash
Neo-- has quit [Ping timeout: 256 seconds]
<lejonet>
infinisil: I think what joepie91 is saying that even tho nodes can claim that they store a certain hash, you can't prove it
<infinisil>
and chances are nobody knows which data belongs to whom
<etu[m]>
joepie91: If you try to read out your data, and know what data to expect (by hash and size or something), and you can successfully read your data. Then *someone* is storing it.
<joepie91>
infinisil: you're misunderstanding the problem
<lejonet>
Unless you fetch the data and can verify that the data you fetched is what it supposed to be, which implies a locally cached copy or so
<joepie91>
etu[m]: yes, and "fetch the entirety of the data every time you want to check" is prohibitively expensive for any kind of automated 'storage assurance' mechanism
<infinisil>
Everytime nodes share data through IPFS, it gets verified that it's actually there, because it needs to have the correct hash on the receiver to be accepted
<joepie91>
especially because it provides no guarantees beyond the instant you've fetched the data
<etu[m]>
But it's the only way to be sure
<joepie91>
infinisil: you're confusing integrity and availability
<joepie91>
IPFS guarantees integrity, not availability
<infinisil>
Yeah, but
<joepie91>
etu[m]: and that's the problem :)
<etu[m]>
Integrity would be nice for upstream sources though
<etu[m]>
:D
<joepie91>
sure, but it's only useful if there's also some degree of availability
<joepie91>
so like
<joepie91>
you can have a dedicated system for seeding this stuff
<joepie91>
but at that point the question is what IPFS is really adding to the mix, if you're trusting that system anyway
<joepie91>
possibly a performance optimization, that *could* help, but at that point it's neither integrity nor availability you're solving
<infinisil>
I can imagine filecoin working by it constantly shuffling around, distributing the data on nodes, so transfer keeps happening -> hash is checked, if it's correct the data has to be there
<joepie91>
infinisil: constantly shuffling the data around is, like previously mentioned, prohibitively expensive
nD5Xjz has quit [Client Quit]
<srhb>
It's a completely nonsensical system. You can, at best, verify at any given time that your data is there, iff your have the data too.
nD5Xjz has joined #nixos
<srhb>
!FUN!
<joepie91>
infinisil: it essentially means you're paying an ongoing (and significant!) traffic cost for data at rest
<lejonet>
srhb: can't have the cookie and eat it too ;)
<joepie91>
infinisil: and that traffic cost is directly proportional to how quickly you want to be aware of data having disappeared
<joepie91>
(in the best-case scenario, that is)
nD5Xjz has quit [Client Quit]
<joepie91>
this also doesn't address proxy attacks, where it's possible that only a single node actually stores your data and all the other nodes are just getting the data *from* that single node and serving it up to you as if they had a copy
<srhb>
lejonet: :'(
<joepie91>
and unless you solve that problem, you have no guarantee that your availability is actually >1, even *if* you verify availability by constantly retrieving data
<lejonet>
srhb: That is the problem hashes mainly try to solve, but then you get into the territory of "How hard is it with collisions?", hashing methodology (how much of the file is hashed, does it include metadata etc etc) and such hard problems :(
<joepie91>
which means that the moment it fails a single availability check, all your data could already be irrevocably gone without any chance of recovery or replication
<joepie91>
tl;dr proof of storage is a currently unsolved problem
<joepie91>
(within the constraint of 'must not require a local copy', that is)
<lejonet>
joepie91: its solved with pifs ;) </sarcasm>
nD5Xjz has joined #nixos
<{^_^}>
[nixpkgs] @dotlambda pushed 2 commits to release-18.03: https://git.io/vxZcu
<{^_^}>
→ 8d24d4fa by @FRidh: python: neovim: 0.2.1 -> 0.2.3
<{^_^}>
→ 8337d1dd by @teto: pythonPackages.neovim: 0.2.3 -> 0.2.4
<joepie91>
lejonet: you joke, but... :(
<lejonet>
srhb: and I just realized I've forgotten to file that PR to fix your issue, sorry!
<lejonet>
joepie91: :P
dan_b has joined #nixos
<srhb>
lejonet: Don't worry about it! I have a number of outstanding issues with it that I haven't had time to even look at yet.
<joepie91>
lejonet: like, I have run into *way* too many people who would genuinely claim "but project X has solved that!", being unable to explain exactly how they did, and basically just deferring to "well they said they did, so clearly you are wrong"
<srhb>
lejonet: (For instance, the mgr assertion becomes an issue if you have an osd-only node, which seems rather unfortunate)
<joepie91>
lejonet: (incidentally, such project-X-es are almost inevitably scummy ICO type deals)
Khetzal has quit [Remote host closed the connection]
<lejonet>
srhb: yeah, the mgr assertion should only be done for mon nodes, even tho its not a hard requirement, it maybe should even be removed
<lejonet>
joepie91: know exactly what you mean there :/
<srhb>
lejonet: I think it should, but yeah. Anyway I'll probably get back to you once I actually have some time to contribute rather than just flail my arms :-P
<lejonet>
srhb: its just highly recommended that each node that has 1 or more mon's should have 1 mgr too
<lejonet>
srhb: :P I'll gladly take your input ^^
<joepie91>
incidentally
<joepie91>
[10:54] <etu[m]> Blockchain for sourcecode-releases :p
<joepie91>
you may be looking for an append-only ledger
<joepie91>
sec
<srhb>
lejonet: I also think I found the issue with ceph-volume, if you didn't. Apparently it's just disabled entirely since the site-packages in $out isn't on PYTHONPATH during the build
<infinisil>
joepie91: Well we don't know what comes out of Filecoin yet, they're still developing it after all, and i don't think they're just spinning thumbs doing no research
<joepie91>
etu[m]: (Chronicle is an append-only ledger)
<lejonet>
srhb: I actually haven't looked at that part yet :)
<srhb>
Consider it a hint :-)
<joepie91>
infinisil: I'm sure, but that means I treat it like any other research project - it hasn't solved anything until it has, and until then it's speculative and may or may not succeed, and therefore it's not worth considering in a practical current-day context when trying to solve a problem
<lejonet>
My guess is you need to add some cmake flag for it to be built or so, similar to cephfs
<srhb>
lejonet: No, it builds by default.
<joepie91>
infinisil: which means that for all practical purposes and intents, IPFS does not in any way ensure availability :)
<lejonet>
srhb: oh, interesting
<lejonet>
srhb: so the build "only" has to be told how to find its own modules (rados.so and friends) for it to be built then?
<srhb>
lejonet: It just warns you that it doesn't anyway if the PYTHONPATH does not contain $out/.../site-...
<srhb>
lejonet: I think so.
kiloreux has joined #nixos
<lejonet>
srhb: I'll add that to my notes file directly so I don't forget it when I actually come around to looking at it
<srhb>
:-)
<infinisil>
joepie91: Yeah, but considering they got so much budget I can imagine them having much more people power than any previous project. Doesn't guarantee success, but the chances are higher. I have my hopes up
Khetzal has joined #nixos
nD5Xjz has quit [Client Quit]
<joepie91>
infinisil: budget is totally irrelevant to me, frankly; right now this looks like an unsolvable problem, so unless the contrary is shown I'm not considering them for practical purposes
<joepie91>
a massive amount of time and money has already been thrown at this topic with no real results
<joepie91>
by many parties
<joepie91>
that doesn't mean it's inherently _impossible_, but I'm not optimistic :)
<infinisil>
joepie91: I only believe it to be unsolvable if you show me a proof of it with clean assumptions and axioms too :)
<joepie91>
emphasis: looks like :)
emilis has joined #nixos
<infinisil>
But generally the speed of progress is proportional to money thrown at it
<infinisil>
more or less
nD5Xjz has joined #nixos
<joepie91>
even if that were true (and I'm not at all convinced that it is), it doesn't really matter if the end result is still a negligible chance of success multiplied by a large amount to still be negligible
<joepie91>
infinisil: and if you just consider the sheer amount of blockchain-y hype projects that are currently alive, thousands of them, "show me a proof that it's unsolvable" becomes an unreasonable proposition for anybody for whom this isn't a full-time job
<joepie91>
so I'm going to apply heuristics, and knowledge of the field, and come to the conclusion that it's not worth paying attention to any proof-of-storage project until they actually show that they can mitigate the known attacks :)
alex`` has quit [Quit: WeeChat 2.0.1]
<infinisil>
Yeah, I'm just gonna see what Filecoin provides once it's out
<infinisil>
A lot of things previously thought to be "impossible" have become possible with the right ideas
<joepie91>
honestly, I feel like "once it's out" is a problematic baseline here in the first place
<joepie91>
it's not some kind of known or understood problem and solution that just needs time to implement and release
<joepie91>
it's a speculative research endeavour that may fail or succeed
<joepie91>
it doesn't so much "come out" as it "produces a result"
<ottidmes>
weird, I am getting "warning: ignoring untrusted substituter 'ssh-ng://nix-ssh@server'", while it is in "trusted-substituters = ssh-ng://nix-ssh@server https://cache.nixos.org" and if I check what will be fetched, it contains a path that is only on that substituter. If I actually continue to run it, it however fails with "error: unexpected end-of-file", with no relevant stuff in the journal of either machine
<joepie91>
there is a very good chance that there will never be a production-ready filecoin release
<lejonet>
Shun the non-believers!
<katona>
Does anyone have trouble with plasma 5 on NixOS 18.03? For some reason there is no keyboard shortcut set for changing windows and if I set alt-Tab it won't save it. I could likely change it from command line, haven't tried it yet but still it's on 18.03 not unstable.. Do I have something strange in my configs (although I removed most of my config files)?
<jetien>
Sorry for asking again, but how do you pin nixpkgs to a fixed version? Is there any documentation for that?
<srhb>
What is this -ng thing?
<srhb>
jetien: I believe there's a wiki article on the subject
* joepie91
stops clogging the channel with blockchainy stuff
<infinisil>
joepie91: Yeah, i meant it more as a once we know more about how they're trying to solve it
mudri has quit [Ping timeout: 264 seconds]
<joepie91>
infinisil: any continuing discussion about this should probably go to #nixos-chat :)
alex`` has joined #nixos
<katona>
jetien: You can clone the GitHub repo and set NIX_PATH variable to "nixpkgs=locationOfGitRepo"
<srhb>
In Nix 2.0 I assume you can just pkgs ? builtins.fetchGit { thatrevision... }
<srhb>
Which is awesome.
<kiloreux>
Assuming I have binaries inside tar.gz and just want to copy them to a specific location and treat them as nix package with no compiling or anything. How can I do that ?
<jetien>
katona: thanks. i saw that somewhere, too. i assumed it's outdated because my $NIX_PATH just contains a plain path. also there is a manifest.nix file
<kiloreux>
and I will be using fetchUrl regularly I assume ?
robstr has joined #nixos
<srhb>
kiloreux: Yup.
<symphorien>
but be aware that non-nix build binaries won't work on nixos, proabaly
pie_ has joined #nixos
<kiloreux>
it's specific for linux. And same architecture also.
<srhb>
That doesn't help
<srhb>
We don't have /lib/
<symphorien>
if they are build with the glibc, you lose
<kiloreux>
Which they are :(
<ottidmes>
kiloreux: You will need to use makeWrapper or patchelf to fix them
<etu[m]>
srhb: But if a person don't want to make a package, just want to run something. It can ;)
<srhb>
etu[m]: Yes, sorry. :P
<kiloreux>
Thank you. What if I want to use older libraries to compile an old package ?
<symphorien>
kiloreux: grepping nixpkgs for "similar" packages may help as well
<etu[m]>
srhb: I agree that patchelf is nicer in every way :)
<kiloreux>
Well the full story is that we need to have support for an older version of ffmpeg.
<kiloreux>
So current nix one is not an option.
<symphorien>
kiloreux: the quick way is to checkout an old nixpkgs, use the ffmpeg of the olf nixpkgs with the new compiler and hope
<srhb>
kiloreux: oldnixpkgs = pkgs.fetchFromGitHub { ... thatrevision of nixpkgs that has older ffmpeg }... oldnixpkgs.mkDerivation ...
<symphorien>
or use the old nixpgks all the way
joelpet has quit [Ping timeout: 240 seconds]
<kiloreux>
I can't have multiple version of nix running can I ?
patrl has joined #nixos
ashgillman has quit [Ping timeout: 264 seconds]
<srhb>
Yes.
<symphorien>
nix of nixpkgs ?
<srhb>
Assuming you mean nixpkgs
<srhb>
nix-daemon, probably not
<srhb>
(Which is the closest interpretation of "running Nix" I can find)
patrl has quit [Client Quit]
ashgillman has joined #nixos
<kiloreux>
Well I mostly meant nix binaries. But I am very interested in a this example oldnixpkgs = pkgs.fetchFromGitHub { ... thatrevision of nixpkgs that has older ffmpeg }... oldnixpkgs.mkDerivation ...
<srhb>
kiloreux: For the purpose of building and installing packages, you can mix and match basically as much as you like.
<srhb>
kiloreux: You won't need it in this case, but yes, nix 1.11 and nix 2.0 can coexist at least some of the way.
<srhb>
But the usual case would be to use the newest nix to build any version of nixpkgs, really.
<symphorien>
nix2 can still evaluate prettyy old nixpkgs
<srhb>
kiloreux: Just fill in the blanks and that becomes a working example with oldnixpkgs.ffmpeg etc.
<srhb>
kiloreux: The usual process (which might be obviated by Nix 2.0) is to bootstrap the older nixpkgs with the newer version, eg. pkgs.fetchGitHub is needed mostly to fetch the older sources.
<ashgillman>
Anyone had much experience in installing Nix without root of late? I had previously used proot, but considering just using ~/nix instead of /nix and building everything
ma27 has joined #nixos
<srhb>
ashgillman: That's trivial with nix 2.0 I thought? Doesn't it take a store location with --store ?
<srhb>
Or NIX_REMOTE=local?root=/home/foo/store
<ashgillman>
srhb: I don't know? I was hoping someone might say something like that
<symphorien>
srhb: it only works with user namespaces
<symphorien>
so old centos users are screwed
<srhb>
symphorien: Oh.
<srhb>
That's good to know, I never looked into the internals
<symphorien>
this is just nix user chroot upstreamed to nix in fact
<ashgillman>
Hmm, interasting. Any docs srhb?
<srhb>
Well, I'm not sure it helps with bootstrapping Nix from nothing
<ashgillman>
It will go on a HPC cluster, not sure what is runs but maybe some kind of RHEL
<ashgillman>
srhb: I have previously followed a similar (older?) version of that, I have a proot installation I can bootstrap from
<ashgillman>
But I have to run many shells as jobs, and managing running proot every time has proven difficult in the past.
<ashgillman>
I think better to spend the time compiling everything, and have it all based from home dir
<ashgillman>
So no little chroot/proot/user-chroot hacks that need to wrap commands
roconnor has quit [Ping timeout: 240 seconds]
<srhb>
I feel like it must be possible to bootstrap nix 2.0 from there and then somehow break out of the chroot/proot/whatever with --store and some mounts.
<srhb>
But I'm not sure how
<guillaum1>
do you know why "fetchGit" may fails (only on travis, not on local) with "warning: rejecte refs/heads/[sha] because shallow roots are not allowed to be updated\nerror: Colud not read [another sha]\nfatal: Failed to traverse parents of commit [yet another sha]"
<manveru>
guillaum1: did you delete ~/.cache/nix/git ?
<guillaum1>
manveru: that's on the travis machine, to be honest I don't have any idea if I have control of this repository.
<ashgillman>
srhb: Yes, was hoping so. I can build from source but also previously had perl version issues then. Might try this route for now
<guillaum1>
manveru: repository -> directory
<manveru>
guillaum1: you should have control over all directories, but i'm not sure what's put into the travis cache by default
<dtz>
guillaum1: did you specify a rev (sha1) as "ref"? Although that error is strange, maybe that's what happens if you do?
<guillaum1>
dtz: no, just `fetchGit ./.;`
ottidmes has quit [Ping timeout: 264 seconds]
rauno has quit [Remote host closed the connection]
<symphorien>
isn't src = ./. enought ?
<dtz>
I guess maybe Travis does a shallow clone? :/
ottidmes has joined #nixos
zzamboni has quit [Ping timeout: 252 seconds]
<dtz>
I like fetchGit ./., nix itself uses it...
<guillaum1>
dtz: haaa, yes, ofcourse, the local clone of travis is shallow. Damned. Well, in the case of travis I may replace fetchGit ./. by ./., but I need a way to detect that I'm on travis. perhaps a derivation argument.
zzamboni has joined #nixos
pkill9 has joined #nixos
<{^_^}>
[nixos-org-configurations] @edolstra pushed to master « hydra-provisioner: Use nixos-17.09-small »: https://git.io/vxZRE
zzamboni has quit [Remote host closed the connection]
<guillaum1>
dtz: actually fetchGit ./. has the issue, but fetchgit {url = ./., sha256 = null;} has not (note the fetch*G*it versus fetch*g*it), this is weird.
nD5Xjz has left #nixos ["WeeChat 2.0"]
<dtz>
fetchGit is the builtin, uhh interesting fetchgit handles this--maybe it's something fetchGit can handle after all?
<guillaum1>
dtz: ho, no sorry, I'm wrong, fetch*g*it fails also (I was confused with my travis report)
<dtz>
haha okay, np
nD5Xjz has joined #nixos
zzamboni has joined #nixos
noam has joined #nixos
vidbina has quit [Ping timeout: 264 seconds]
orivej has joined #nixos
leothrix has quit [Ping timeout: 240 seconds]
davidlt_ has quit [Remote host closed the connection]
zzamboni has quit [Ping timeout: 256 seconds]
noam has quit [Read error: Connection reset by peer]
<nioncode>
I run nix on my work computer and am facing a problem: I have to compile some source code for an embedded system for which I only have a prebuilt gcc based toolchain, which obviously does not work at all on nix (starting with having the wrong dynamic interpreter set). is there a clean way to get this toolchain up and running inside a nix-shell environment? or do I really have to patchelf all those binaries in
<nioncode>
the toolchain and probably create wrappers for each of them to set the correct LD_LIBRARY_PATHS?
<gchristensen>
nioncode: you could make an fhs environment
<nioncode>
gchristensen: thanks, that looks interesting. will try this asap
<disasm>
zimbatm: yeah, the error is output in readShebang before it starts parsing the rest of the file. So it checks if line 1 is a valid shebang (which isn't quite accurate, it checks that line 1 has a the form of #!something), if it doesn't (e.g. comment or blank line) it skips any comment or whitespace lines looking for a shebang, it finds one it use the shell listed there but throws a different warning that first
davidlt_ has quit [Remote host closed the connection]
<__monty__>
gchristensen: Haven't been able to run the index command to build the database because of restricted ram.
aloiscochard has joined #nixos
<gchristensen>
ouch
logzet has joined #nixos
<joepie91>
nioncode: gcc is GPL, isn't it?
<joepie91>
nioncode: wouldn't that make it a license violation to only provide build artifacts?
<joepie91>
(ie. prebuilt toolchain)
<joepie91>
ie. possibly ask whoever provided it to give you the source code :)
davidlt has joined #nixos
<Yaniel>
not if it's unmodified I think?
<Yaniel>
since then whoever gives you the binary can point you at the official tarball
<nioncode>
joepie91: might be. but companies usually don't really care, since nobody sues them anyway
<joepie91>
Yaniel: sure, but that'd have the same practical effect
<joepie91>
nioncode: worth a shot
<gchristensen>
nioncode: good news, one of our founding community members spends his life suing over GPL violationss
<nioncode>
gchristensen: thanks, buildFHSUserEnv worked like a charm
<gchristensen>
cool :)
lord| has quit [Quit: WeeChat 2.0.1]
<nioncode>
gchristensen: I don't think I want to sue our kind of partner for such a thing. they would probably send me the source if I ask, but since buildFHSUserEnv works so nicely, I don't care much
<{^_^}>
[nixpkgs] @jtojnar pushed to master « haskell-configuration-common: disable some tests depending on tasty-hunit-0.9 »: https://git.io/vxZwE
<Myrl-saki>
scErr
<Myrl-saki>
Is there an "error" in Nix?
ashgillman has quit [Ping timeout: 264 seconds]
<Myrl-saki>
Oh, abort.
etiennebatise has joined #nixos
emilis has joined #nixos
<Myrl-saki>
Wtf. I actually checked my code, and it already says "abort."
<Myrl-saki>
The subconscious~
<fendor>
Hey, I have a question to haskell development using nixos and stack. Currently I have to add - - nix to every stack invocation, or add enable: true in stack.yaml. The latter is generally not usable because collaborating with people who don't have nix can't compile the project anymore. But I can only use the current haskell-ide-engine if it is enabled in the yaml, otherwise it tries to download a whole new ghc. Can somebody explain what the
<fendor>
preferred way of haskell and nix development is?
<Myrl-saki>
I removed it, I should have replaced it with an empty ""
<ottidmes>
Anybody else experienced "unexpected end-of-file" with SSH substituters in Nix 2.0 with on 17.09 stable? I am running Nix 2.0 and its daemon on all relevant machines
<Myrl-saki>
I was rebuilding ghc lmao
ryanartecona has joined #nixos
vidbina has joined #nixos
<{^_^}>
[nixpkgs] @edolstra pushed to master « zstd: Fix install name on macOS »: https://git.io/vxZrR
<clever>
7449 store paths deleted, 7971.25 MiB freed
<clever>
according to datadog, it froze at 23:40
<clever>
and it is now 10am
<disasm>
gchristensen: how do you manage your zsh config? Do you do that in your home directory? git grepping your nixos-configs I found a very minimal setup
<disasm>
Mainly asking because 18.03 broke a lot of things in prezto and looking at redoing mine.
mightybyte has joined #nixos
<gchristensen>
disasm: tbh I have almost no zsh config
nuncanada has joined #nixos
<disasm>
gchristensen: oh :) minimalist, eh?
<gchristensen>
disasm: I used to use ohmyzsh but got frustrated when it'd slow down my prompt, so I deleted it and I think I forgot to set anything back up again :$
<clever>
#6 0x556519b757a5 [17670:17703:0320/101623.355468:ERROR:socket_posix.cc(91)] CreatePlatformSocket() returned an error, errno=24: Too many open files (24)
<clever>
gchristensen: and now chrome refuses to launch
blankhart has quit [Ping timeout: 256 seconds]
<gchristensen>
ruh roh
Guanin has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @dtzWill opened pull request #37447 → python*: set thread stack size on musl → https://git.io/vxZiB
coot has quit [Quit: coot]
<clever>
security.pam.loginLimits time...
<Dezgeg>
IIRC it's not safe to increase the default limits from 1024
<clever>
ive been increasing it to 4096 for months, for just chrome
<Dezgeg>
since the file descriptor bitmaps used by select() have a hardcoded size
<{^_^}>
→ 45903094 by @ttuegel: Merge pull request #36868 from bkchr/kde_frameworks_5_44
<clever>
Limit Soft Limit Hard Limit Units
<Dezgeg>
I guess if you limit it to just chromium it should be safe though
<clever>
Max open files 8192 65535 files
<clever>
i set the limit to 64k,and chrome lowered itself to 8k
<Dezgeg>
well, assuming chrome doesn't use select :P
<clever>
i have had problems in the past, that i tracked down to a single renderer process opening so many sockets that it commited suicide
<clever>
but its never affected the master process before
<catern>
has anyone ever investigated an alternative representation of package source versions in nixpkgs, by using git submodules at specific revisions to represent the source?
<catern>
with fetchgit, this would actually be optimized and fast now
<clever>
[root@amd-nixos:~]# ls -l /proc/445/fd/ | wc -l
<catern>
slightly more detail: you'd have a repo with a bunch of submodules, one submodule per package, and the source code version for some package A would be obtained by doing a fetchgit on the submodule for package A
<{^_^}>
→ 06828573 by @dtzWill: Merge pull request #37447 from dtzWill/fix/python-musl-thread-stacksize
<YegorTimoshenko[>
clever: thanks, NIX_CONF_DIR can be used for that purpose, but i only want to change one option. i probably could generate nix.conf based on /etc/nix/nix.conf at eval time...
<infinisil>
srid: I pretty much only needed the pulseaudioFull and then connect with bluetoothctl
zzamboni has joined #nixos
<infinisil>
srid: and i sometimes have to `systemctl restart bluetooth` and reconnect with bluetoothctl
<silver_hook>
Is it “normal” (for NixOS) that when in Plasma I install a theme pack through GHNS, the dialog says it’s installed, but the content doesn’t show in the settings?
<gchristensen>
clever: how did you send that over pushbullet?
endformationage has joined #nixos
jetien has quit [Quit: Leaving.]
<katona>
That's the reason there is so many failing jobs :0 I couldn't find any issue on a quick try, is it a known problem?
humanoyd has joined #nixos
<{^_^}>
[nixpkgs] @Ericson2314 merged pull request #37373 → Add space for NIX_CXXSTDLIB_COMPILE+=' ...' as NIX_CXXSTDLIB_COMPILE might not be empty. → https://git.io/vxs8T
<quez>
I have a local package (with a nixpkgs pinned by git hash) that builds on MacOS. When I try to build it on NixOS, I get a "No such file or directory" complaining about one of my intermediate build products and indeed that file is absent. Does anyone have an immediate intuition for what could be going wrong?
<kiloreux>
I have a static build of ffmpeg in a tar and trying to create a nix package with it. However I get the following mv: target '/nix/store/g1hr75lb6c9srmp0vxab40lnm1rim700-ffmpeg' is not a directory
<kiloreux>
builder for '/nix/store/4nh2cvd8c6iwsswpaa9ky2ac7cf06mrb-ffmpeg.drv' failed with exit code 1
<coconnor>
quez: Mac is "case preserving" but not case sensitive. Which can lead to fun iisues
<quez>
I can see Nix building the missing file at one point. It seems like it gets cleaned up prematurely or something
<sphalerite>
kiloreux: create it first :)
<sphalerite>
kiloreux: make your installPhase mkdir $out ; mv * $out
<quez>
i.e. I see `building path(s) ‘/nix/store/znsss28yqylxdyaa4f85bk8fh084klpr-package.json’` at one point, a separate `build path ...` and then a `getting attributes of path ‘/nix/nix/store/znsss28yqylxdyaa4f85bk8fh084klpr-package.json’: No such file or directory`
<kiloreux>
Thank you sphalerite
<kiloreux>
That solved it.
<{^_^}>
[nixpkgs] @zimbatm opened pull request #37459 → nix-serve: patch for nix 2.0 → https://git.io/vxnkk
rauno has quit [Ping timeout: 245 seconds]
Arcaelyx has joined #nixos
Slabity has joined #nixos
__monty__ has quit [Read error: Connection reset by peer]
<tnks>
are nix-shell interpretter scripts going anywhere with Nix 2.0?
<tnks>
As far as I can tell, the new "nix" command doesn't subsume it... but I'm not sure how to Google for sentiments on that.
<Slabity>
Using the example in the nixpkgs manual results in the same issue
<coconnor>
Slabity: remove the "inherit pkgs".
<tnks>
I have a feeling there are too many ways to "run" a Nix command, and I like that "nix run" unifies a lot of them.
<coconnor>
Slabity: I'd also remove the "with pkgs;"
<tnks>
gchristensen: okay, good to know it's not on the cutting board yet.
<Slabity>
coconnor: I made both changes, but the issue is still there.
JosW has joined #nixos
<Slabity>
Wait a second... I may have fixed it?
<rawtaz>
hmm. so this "previous versions" thingie, how does that work anyway. lets say that from version A to version B i removed almost all my software and installed a lot of new software - is the stuff for version A still on disk so i can fire that version up completely offline and even if i cleaned all caches etc there is, or will they have to be installed anew when loading version A (e.g. requiring internet access)?
richard has joined #nixos
<Slabity>
coconnor: Nevermind, it was just a dumb typo in my higher config. Thanks anyways
zzamboni has joined #nixos
richard is now known as Guest48926
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #37460 → treewide: `meta.platforms` fixes now that it tracks the host platform → https://git.io/vxnLL
silver has joined #nixos
rauno has joined #nixos
silver has quit [Remote host closed the connection]
zzamboni has quit [Ping timeout: 246 seconds]
<wilornel>
Hey #nixos! Does anyone know of a nice solution for user defined long running interactive processes that should ideally always be running?
<vdemeester`>
I have some weird terminal display issue running `nixos-rebuild -I nixpkgs=https://nixos.org/channels/nixos-18.03/nixexprs.tar.xz -I nixos-config=./myconfiguration.nix build-vm` on a 17.09 nixos system.. is it expected ?
zzamboni has joined #nixos
xAFFE has left #nixos ["Error from remote client"]
<{^_^}>
[nixpkgs] @svanderburg pushed to master « titaniumenv: add 7.1 SDK and make it the default »: https://git.io/vxntP
zzamboni has quit [Remote host closed the connection]
asuryawanshi has joined #nixos
zzamboni has joined #nixos
<coconnor>
wilornel: if you need a long running *interactive* process then screen/tmux plus loginctl solinger is likely required
Olgierd has joined #nixos
Olgierd has quit [Changing host]
Olgierd has joined #nixos
<coconnor>
s/solinger/enable-linger/
<coconnor>
wilornel: tho there are some fancy "attach seat" commands in loginctl that might be relevant
kthnnlg has joined #nixos
orivej has quit [Ping timeout: 256 seconds]
strleok7 has quit [Remote host closed the connection]
<coconnor>
emphasis on the interactive because if a TTY is not required then a simple systemd service would work
<catern>
hmmm
zzamboni has quit [Ping timeout: 264 seconds]
<catern>
hey #nixos, does anyone have any clever ideas about how I might, in a Nix expression, running in a git repository, determine the commit hash of a submodule?
<clever>
coconnor: where was linger set again?, ive noticed systemd killing my screen sessions
<catern>
the easiest way is to shell out to git, but that's pretty impure
<catern>
(well I mean, it's not actually impure, but anyway)
<clever>
coconnor: builtins.fetchGit for 2.0
aminechikhaoui has quit [Read error: Connection reset by peer]
<quez>
Ah, figured out there was a double `/nix/nix` in the path due to a bad relative path
<catern>
clever: i assume that was directed at me
<clever>
catern: lib.commitIdFromGitRepo for any nix version
<catern>
does fetchGit actually work with submodules? keep in mind I don't want to know the commit of a git repo, I want to know the commit of a submodule inside a git repo
<{^_^}>
[nix] @dtzWill opened pull request #1997 → ask autotools for c++14 support flags, not c++11; don't override later → https://git.io/vxnqx
<catern>
clever: woah, you mean like, if repo root is repo/, and submodule is at repo/submodule, I could just do fetchgit "repo/submodule"?
<tnks>
hmm... #!nix-shell invocations are confusing. I can just specify Haskell packages with -p and they get woven into the interpretter specified with -i?
<robstr>
Hey, first time nixops user, when running `nixops deploy ...` i receive `at /nix/var/nix/profiles/per-user/root/channels/nixos/nixpkgs/lib/modules.nix:109:60`
<clever>
catern: all it does is impurely run some git commands at eval time
<catern>
oh yeah, that's true
<catern>
so I guess it depends on whether git fetch supports that
<kthnnlg>
Hi, I'm using nixos on my mac, and am trying to build a package that I need to compile with gcc. However, the configure script of the package picks up clang instead of gcc. Does anyone know how I might configure my nix-shell instance or the build script to prefer gcc?
<clever>
kthnnlg: patch the build scripts to obey $CC and $CXX
<kthnnlg>
ok, i think they do
<clever>
nix-shell already points those to clang
zzamboni has quit [Ping timeout: 256 seconds]
<kthnnlg>
clever: yep, it seems so. is there a way for me to force nix-shell to point to gcc?
<clever>
not sure how to do that on darwin
sdhand has quit [Ping timeout: 265 seconds]
Taneb has quit [Ping timeout: 246 seconds]
<tnks>
wait, is the argument of "-i" for nix-shell not a normal Nixpkgs derivation?
<{^_^}>
[nixpkgs] @zimbatm pushed to master « nix-serve: patch for nix 2.0 (#37459) »: https://git.io/vxnmF
Taneb has joined #nixos
<{^_^}>
[nixpkgs] @zimbatm pushed to release-18.03 « nix-serve: patch for nix 2.0 (#37459) »: https://git.io/vxnYT
<{^_^}>
[nixpkgs] @Twey opened pull request #37462 → bitmath: init at 1.3.1.2 → https://git.io/vxnYk
aminechikhaoui has joined #nixos
<mojjo>
hi! how can I declare in a nixos configuation, that a package should be installed in an older version? particularily I have some issue with the latest chromium (65), it always crashes when I copy sth to the clipboard. The 63 version works but how can I make this permanent?
<catern>
clever: oh, but fetchGit can't fetch uncommitted changes, can it?
<{^_^}>
[nixpkgs] @uvNikita opened pull request #37463 → todolist: init at 0.8 → https://git.io/vxnYC
zzamboni has quit [Ping timeout: 252 seconds]
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #37464 → lib: Make platform predicates greppable history fix → https://git.io/vxnYz
aminechikhaoui has quit [Read error: Connection reset by peer]
<{^_^}>
[nixpkgs] @Ericson2314 merged pull request #37464 → lib: Make platform predicates greppable history fix → https://git.io/vxnYz
<{^_^}>
→ 175d4ab1 by @Ericson2314: lib: Make platform predicates greppable
<{^_^}>
→ 2ce3df3f by @Ericson2314: Merge commit '543210751f46a4787fd0a5ef4c3bfa9a4b789e3c' into greppable
<{^_^}>
→ 326c9f50 by @Ericson2314: Merge pull request #37464 from obsidiansystems/greppable
zzamboni has joined #nixos
<rawtaz>
mojjo: in an attempt to help, even though i dont know if there's a better way, here's how i specify a *newer* version of a package (newer than that's in the unstable channel): https://pastebin.mozilla.org/9080461
vidbina has joined #nixos
<tnks>
Actually... I think my suspicions of nix-shell are warranted. The documented way of using Haskell packages with it doesn't work.
<{^_^}>
[nixpkgs] @timor opened pull request #37465 → workcraft: init at 3.1.9 → https://git.io/vxnYA
boomshroom has joined #nixos
aminechikhaoui has joined #nixos
quez has quit [Ping timeout: 245 seconds]
<boomshroom>
I have to go soon, so I'd like to make this quick. How can I have a derivation that creates a file in a specific directory of its output? Ie toFile "/etc/blah" "..."?
<kthnnlg>
Hi All, I'm using this guide https://blog.galowicz.de/2018/02/27/managing_libraries_with_nix/ as a reference for how to force nix-shell to use gcc instead of clang on darwin. However, what I see is the following `$ nix-shell --pure -p gcc7 --arg stdenv "with (import <nixpkgs> {}); overrideCC gccStdenv gcc7" --command "\$CXX --version" clang version 5.0.1 (tags/RELEASE_501/final)`. Does anyone know how I might get the desire
spietz has quit [Quit: WeeChat 1.9.1]
aminechikhaoui has quit [Read error: Connection reset by peer]
<mojjo>
rawtaz: thanks for the code.. However, I think I see now that I have another problem. even if I install the 63 version via nix-env the latest version is opened. (It's linked to the older one but inside the settings of the browser the latest version shows up)
zzamboni has joined #nixos
<boomshroom>
Basically, I'm setting i3 to look in my nix-profile so I can generate it with nix (without one of the many 3rd party home managers).
<tnks>
also, "runghc" isn't a package in the root of "nixpkgs", so how does nix-shell find it as an interpretter?
<rawtaz>
mojjo: uh, where is it getting the latest version from? is it just some value in your configuration, or is taht latest version still installed somewhere?
<tnks>
oh... it's just a command called... sorry, I get it.
aminechikhaoui has joined #nixos
Guest91821 has quit [Changing host]
Guest91821 has joined #nixos
Guest91821 is now known as sdhand
<rawtaz>
boomshroom: good question, i tried to find the answer real quick, no dice.
<boomshroom>
I know from a previous thing I was trying that there's so many ways to do the same thing.
<wilornel>
coconnor: It would be nice to discuss that solution later
<wilornel>
which is not at all what I was asking about but I find it interesting
<clever>
catern: there is an open bug with builtins.fetchGit, it will report a hash of all 0's if there are any un-commited changes
<clever>
catern: it will also apply a filterSource over the whole tree, to delete anything git isnt aware of
<rawtaz>
boomshroom: perhaps you can take one of the common things such as sshd and check how it generates the config files under /etc
<boomshroom>
On an unrelated note, I'm tempted to try a sort of "NixOS From Scratch" inspired by LFS, except the builds are all done with Nix. One rule: No nixpkgs outside of bootstrapping gcc.
<catern>
clever: ah. sounds unsuitable...
<mojjo>
rawtaz: ok, I removed chromium from the global config.. And yeah, the command does not exist anymore. (also not in root's env).. When I open `/nix/store/vbhl1xx92xcsxb3f8iza5fdxppiv5fsg-chromium-63.0.3239.108/bin/chromium` the browser settings say 65
zzamboni has quit [Ping timeout: 248 seconds]
aminechikhaoui has quit [Read error: Connection reset by peer]
<rawtaz>
mojjo: does that mean the problem is solved?
<rawtaz>
mojjo: also, what did you do to specify which version you want?
<clever>
boomshroom: go hard, bootstrap gcc as well, its easy!
<mojjo>
rawtaz: no, the contrary
<maurer>
mojjo: Is the browser already running elsewhere? Chrome/chromium apps first look for a running instance of chrome/chromium and spawn a new window off the main process
<rawtaz>
mojjo: right you want 63
<boomshroom>
clever: I mean. I can't compile GCC without GCC. The only ways to do that is to pass in a GCC from nixpkgs, or look at the path.
<{^_^}>
→ c234c01d by @kmicklas: zoom-us: add desktop item
<{^_^}>
→ ec8c3c82 by @Ericson2314: Merge pull request #37190 from kmicklas/zoom-us-desktop-item
<{^_^}>
→ 0e9d1fad by @kmicklas: zoom-us: fix preInstall
<{^_^}>
[nixpkgs] @Twey opened pull request #37469 → toml: init at 0.9.4 → https://git.io/vxnGL
<nliadm>
so with nix 2.0, pointing nix-shell at a file that evaluates to a derivation now fails if the derivation doesn't actually build
<{^_^}>
[nixpkgs] @Ericson2314 merged pull request #37460 → treewide: `meta.platforms` fixes now that it tracks the host platform → https://git.io/vxnLL
<{^_^}>
→ e38d2884 by @ryantm: lua51Packages.lgi: 0.9.1 -> 0.9.2
<{^_^}>
→ fceb73de by @rasendubi: Merge pull request #37468 from ryantm/auto-update/lgi
ma27 has joined #nixos
maximiliantagher has joined #nixos
raynold has quit [Quit: Connection closed for inactivity]
aminechikhaoui has quit [Read error: Connection reset by peer]
boxofrox has joined #nixos
aminechikhaoui has joined #nixos
rauno has joined #nixos
jensens has quit [Ping timeout: 240 seconds]
<{^_^}>
[nixpkgs] @svanderburg pushed 2 commits to release-17.09: https://git.io/vxnZD
<{^_^}>
→ 8968f683 by @svanderburg: titaniumenv: add 7.1 SDK and make it the default
<{^_^}>
→ 4fcf69f9 by @svanderburg: titaniumenv: remove obsolete java fix wrapper on darwin
coot has joined #nixos
<acowley_>
If I do nix-shell -p llvmPackages.libcxxClang on a NixOS machine, I still get glibc on the compiler's search path. Is there any way to avoid that?
<coconnor>
pff I can't remember how to ask the bot to exec a nix statement, but
<coconnor>
let x = /proc/cpuinfo ; y = "${x}" ; in toString y
<gchristensen>
#nixos doesn't have such a bot
<coconnor>
OH
<infinisil>
Damnit
<boomshroom>
I can I write an IRC bot? :P
<infinisil>
I tried to create one once, but nix-repl broke gchristensen's bot completely, and hnix doesn't work well
<infinisil>
(as in i wrote a bot backend for gchristensen's frontend)
jrolfs has quit [Quit: WeeChat 2.0]
<infinisil>
boomshroom: If you have the time please do!
dahirsch has joined #nixos
<boomshroom>
So you already tried to write a nix-evaluator irc bot?
<infinisil>
boomshroom: A nix-repl bot would be so nice
mudri has joined #nixos
<infinisil>
boomshroom: The way to go is to incrementally define a file and evaluate that using nix-instantiate
<{^_^}>
[nix] @dtzWill opened pull request #1998 → execl: cast NULL sentinel to (char *), per man page and compiler warning → https://git.io/vxnzY
<infinisil>
so e.g. start with `let \n\n in null`, then when somebody types `x = 10`, make the file `let \n x = 10; \n in null`
<dahirsch>
question concerning uefi: im using grub2. netbook does not boot with 'uefi only' option. it does boot with 'legacy enabled' and if i choose 'uefi first' or 'legacy first' is irrelevant. when it does boot, it uses uefi (tested by inspecting /sys). any idea how to use it with 'uefi only'?
<infinisil>
then when somebody types `x`, make the file `let \n x = 10; \n in x`
<infinisil>
which will output the value of x, which you can then print
<infinisil>
dahirsch: have you set boot.loader.grub.device = "nodev"?
<{^_^}>
→ d16bf733 by @hedning: gnome3.gnome-shell: add NetworkManagerApplet
<{^_^}>
→ 4cdb32e5 by @jtojnar: gtk3: fix gtk-encode-symbolic-svg
<{^_^}>
→ e1e41af5 by @jtojnar: webkitgtk220x: cleanup
<boomshroom>
Unrelatedly: I ended up searching emoji. Apparently there are wizard and fairy emoji and unicode 11.0 will be adding superheroes and redheads.
<boomshroom>
What kind of fonts have you guys been using for emoji?
<coconnor>
boomshroom: check the source to the "nix" command. Pretty clean c++ demonstrating using the library for a repl
<catern>
infinisil: oh wow, okay, perfect start :) but that leaves the location of the nixpkgs repo as a prefix, so next question: is there a way to remove a path prefix from a path?
<infinisil>
catern: baseNameOf?
acowley_ is now known as acowley
<infinisil>
And paths are just strings, so you can just use the string manipulation functions
<clever>
paths are a bit weird though
<clever>
./. + "/" == ./.
<catern>
infinisil: I don't want the basename, I want to take "a/b" and "a/b/c/d" and get "c/d"
vidbina has quit [Quit: leaving]
<infinisil>
catern: "a/b" + "/" + "c/d" :)
<clever>
also, "${./.}" is very different from (toString ./.)
<clever>
infinisil: and ./. + "/" + "c/d" will fail horridly
FareTower has quit [Ping timeout: 248 seconds]
<catern>
infinisil: I think I explained poorly, the inputs are "a/b" and "a/b/c/d"
<catern>
"c/d" is the output
Synthetica has quit [Quit: Connection closed for inactivity]
<clever>
infinisil: nix strips the trailing / off after each append
<infinisil>
clever: Yeah you told me before too :) They are *almost* string
<infinisil>
oh
<infinisil>
catern: removePrefix i think
<catern>
more concretely, my inputs are "/home/sbaugh/.local/src/nixpkgs" "/home/sbaugh/.local/src/nixpkgs/sources/tools/system" and I want "sources/tools/system"
<catern>
okay, so then the final thing is, is there any variable in nixpkgs.lib already which contains the location of the nixpkgs root?
<catern>
so I can do the relative thing without having to rely on removePrefix ../../.. ./.
mudri has quit [Ping timeout: 248 seconds]
<catern>
and can just do removePrefix nixpkgsroot ./.
<infinisil>
dahirsch: Maybe clever could help with this, he's pretty experienced with Bios/Uefi and boot problems
<infinisil>
catern: <nixpkgs>
<catern>
infinisil: that will fail whenever NIX_PATH isn't set
<dahirsch>
okay, ill message him
<infinisil>
catern: Well what is the base if you don't know a <nixpkgs>?
<infinisil>
catern: But you can rely on the depth of the file never changing
<catern>
it's ../../.. if i'm three levels deep :)
<infinisil>
catern: lots of places in nixpkgs do that
<{^_^}>
[nixpkgs] @dtzWill pushed 5 commits to release-18.03: https://git.io/vxn2Q
<{^_^}>
→ 828d15e9 by @dtzWill: python*: set thread stack size on musl
<{^_^}>
→ 67cdc01b by @dtzWill: llvm4: same treatment
<{^_^}>
→ 043ee6c6 by @dtzWill: llvm_5: patch sanitizers to enable w/musl
<clever>
catern: pkgs.path is the path to the nixpkgs that pkgs came from
<clever>
dahirsch: *waves*
<infinisil>
dahirsch: (just ask here)
<dahirsch>
ok then
<infinisil>
clever: Oh i didn't know that
<infinisil>
Nice
<clever>
dahirsch: that config looks like it should work, what is failing?
<dahirsch>
clever: question concerning uefi: im using grub2. netbook does not boot with 'uefi only' option. it does boot with 'legacy enabled' and if i choose 'uefi first' or 'legacy first' is irrelevant. when it does boot, it uses uefi (tested by inspecting /sys). any idea how to use it with 'uefi only'?
<{^_^}>
→ ab67f36e by @abbradar: xdg_utils: fix xdg-screensaver
<{^_^}>
→ 79216189 by @abbradar: steam: add dependencies for Surviving Mars
<{^_^}>
[nixpkgs] @abbradar pushed 2 commits to release-18.03: https://git.io/vxnrc
<{^_^}>
→ 69e8d7dd by @abbradar: steam: add dependencies for Surviving Mars
<{^_^}>
→ 97a2621d by @abbradar: xdg_utils: fix xdg-screensaver
<catern>
clever: infinisil: but how can I construct this path programmatically and have it still automatically pull things into the store when used?
<dahirsch>
clever: nixos-rebuild switch does not trigger the reinstallation of grub?
<clever>
dahirsch: one sec
<clever>
dahirsch: in the nixos-rebuild man page is --install-bootloader
ma27 has joined #nixos
<thoughtpolice>
Does the new nix run take an -I flag? Where does it get nixpkgs from on NixOS?
<clever>
thoughtpolice: it still obeys NIX_PATH and -I as always
<{^_^}>
[nixpkgs] @Ericson2314 pushed commit from @angerman to master « haskell-generic-builder: Add `enableHsc2hsViaAsm` for Windows cross compilation »: https://git.io/vxnrK
<{^_^}>
[nixpkgs] @Ericson2314 merged pull request #37479 → haskell-generic-builder: Add `enableHsc2hsViaAsm` for Windows cross compilation → https://git.io/vxnaT
<thoughtpolice>
Another weirdness: nix-env -iA and e.g. 'nix run' take different prefix names? nix-env -iA expects e.g. 'nixos.awscli' while nix run uses 'nixpkgs.awscli'
kuznero has joined #nixos
<kuznero>
Hi All!
<clever>
thoughtpolice: ah, i hadnt figured out the prefixes with run and had been using `nix run -f '<nixpkgs>' hello`
<kuznero>
Does ghc exists in Nix packages that is built against musl instead of glibc?
<clever>
kuznero: i dont think so, but Sonarpulse might know more about how to flip that switch
<Sonarpulse>
kuznero: use a cross or native musl stdenv and try it
<Sonarpulse>
no idea if anything special is needed
<Sonarpulse>
domenkozar I think was looking into this
<kuznero>
Sonarpulse: but is it practically possible or just theoretical?
<dtz>
I don't think we have ghc musl support, IIRC missing bootstrap compilers? maybe?
<dtz>
haven't tried at all, just reporting what I remember :)
<kuznero>
:(
<clever>
the cross stuff would be simpler, it can be a glibc based ghc, that targets musl based executables
<clever>
that feels like something easier to manage
<kuznero>
I am dreaming of beating GoLang by compiling Haskell into really statically linked binary
darlan has joined #nixos
darlan has quit [Client Quit]
<catern>
wait, I see how, to do what I want, I can just do: /. + mypath
<catern>
that's hacky but alright
<dtz>
if you do much with musl I strongly recommend using cache.allvm.org, since I don't think cache.nixos.org has any musl-native binaries (err it has things cross-built against musl...)
<dtz>
haha
<clever>
catern: i think builtins.toPath does that already
<catern>
no, builtins.toPath is broken, apparently
<catern>
and does nothing
hph^ has joined #nixos
<kuznero>
Is the story of GHC+Musl better on any other linux distro by any chance?
<thoughtpolice>
clever: I am playing around more and the semantics totally confuse me. Ugh. I figured out what I need though, thanks as usual!
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #37482 → cross stdenv: Make depsBuildBuild overrideable by config too. → https://git.io/vxnoc
<thoughtpolice>
(Mainly: I just wanted to make sure I was always using my /etc/nixos/nixpkgs everywhere, since I try hard to put everything on an upstream-able track)
<clever>
infact, just an ls -ltrh on auto reveals this
<tnks>
clever: what cleans up the broken links in gcroots/auto?
<tnks>
only a nix-collect-garbage?
freeman42x]NixOS has quit [Quit: Leaving]
<clever>
tnks: any form of garbage collection
<clever>
thats the very first step it does
<tnks>
clever: okay, that's good that it creates an indirect link.
<dtz>
(nevermind me, I'm just wrong, 1.0.3 was just released hehe. I probably am thinking of 1.0.2 and since that's what we use already fixed whatever it was xD xD)
<tnks>
but does this mean that if I never do an explicit garbage collection, broken indirect links from every invocation of nix-shell will just accrue?
<tnks>
I'm just not doing any GC on this cluster, because it's Docker-based, and I'm playing games with links.
<pxc>
shlevy: I would like to ask some questions about it and to sound out some propositions, but I'm afraid if I make a PR or an RFC, I'll accidentally write a wall of text and no one will read it
<pxc>
maybe I should just go for it
<shlevy>
pxc: If you make an RFC it will get reviewed and eventually decided on now
fendor has joined #nixos
simendsjo has quit [Quit: ERC (IRC client for Emacs 25.3.1)]
<wervenyt[m]>
anyone use kde connect on nixos?
mudri has joined #nixos
<pxc>
shlevy: I'll write one up after work, then. thanks for the nudge
<MichaelRaskin>
There are multiple versions of what-sh-to-use-in-sandbox
<pxc>
who would win? a programming language used to build massively scalable soft real-time systems with requirements on high availability... or one curly boi?
<MichaelRaskin>
(which is about the Nix build)
<infinisil>
pxc: haha
<infinisil>
MichaelRaskin: I don't get that
<MichaelRaskin>
_anything_ that triggers a rebuild of some packages (erlang included) on a box with unfortunate /bin/sh replacement triggers a failure
<gchristensen>
pxc: obviously one curly boi
<MichaelRaskin>
curl included
<infinisil>
MichaelRaskin: No, I can reproduce it locally
<MichaelRaskin>
Maybe you also have a build of Nix that puts unsuitable /bin/sh inside sandbox
<MichaelRaskin>
At some point there was busybox stripped down below POSIX requirements
katyucha has quit [Ping timeout: 264 seconds]
<freeman42x>
could someone recommend an external PDF viewer that does not render blurry (evince issue)?
<MichaelRaskin>
I guess blurriness depends on PDF
<MichaelRaskin>
zathura uses a different engine
<infinisil>
MichaelRaskin: Are you saying my /bin/sh is bad?
<MichaelRaskin>
Or xpdf
<MichaelRaskin>
infinisil: no
<pxc>
freeman42x: is it a DPI scaling issue? Okular looked horrible on HiDPI until a recent release. It turned out it was rendering at the wrong size
<MichaelRaskin>
I am saying that your Nix puts an over-weakened version of busybox inside the build environment as /bin.sh
<pxc>
anyway Okular is pretty good. It will pull in some KDE deps, ofc
<freeman42x>
pxc: all I can say is that it looks blurry, and it is not caused by Font sub-pixel order
<infinisil>
But this should be ported to 18.03
<freeman42x>
pxc: I installed mupdf but it won't even launch, the executable does something then... nothing happens
<pxc>
freeman42x: the issue I encountered was not a subpixel rendering one. The application misunderstood the desktop resolution, so it rendered the PDF at way, way low-resolution. And then it was scaled up to fill the window, resulting in a blurry experience
<fendor>
Morning, when using criterion on nixos and stack and try to benchmark via `stack bench` it fails with `ffp-reloaded-benchmarks: <stdout>: commitBuffer: invalid argument (invalid character)`. did i miss some libraries that I have to add to stack.yaml?
<freeman42x>
pxc: that could be it, I am running NixOS inside VMWare VM
<pxc>
I found a similar bug for Evince, but it looks like it was fixed years ago
leothrix has joined #nixos
abrar_ has quit [Quit: WeeChat 2.0]
pie_ has joined #nixos
abrar has joined #nixos
<pxc>
gchristensen: what's the impending inode crisis?
<gchristensen>
the "I have no more inodes left" problem
<{^_^}>
→ f9dda86b by @flokli: sqldeveloper: 17.4.0.355.2349 -> 17.4.1.054.0712
<{^_^}>
→ 1c810f5f by @Mic92: Merge pull request #37346 from flokli/sqldeveloper
<pxc>
does anyone else think NixOS could be useful as a basis for more traditional, or even BSD-like distros?
jtojnar has joined #nixos
<wervenyt[m]>
nixos is pretty not-unix
<wervenyt[m]>
and not necessarily in a bad way
<{^_^}>
[nixpkgs] @Mic92 pushed 2 commits to release-18.03: https://git.io/vxnFZ
<{^_^}>
→ 9cd1122d by @jfrankenau: masterpdfeditor: 4.3.61 -> 4.3.82
<{^_^}>
→ 032a3286 by @flokli: sqldeveloper: 17.4.0.355.2349 -> 17.4.1.054.0712
asuryawanshi has joined #nixos
<{^_^}>
[nixpkgs] @ysndr opened pull request #37494 → krunner-pass: init at version v1.3.0 → https://git.io/vxnFn
<pxc>
you could have stable releases of a small system also get linked into something like /StaleOS/1.0/lib /StaleOS/1.0/bin, etc.
<MichaelRaskin>
Does SUS actually include FHS now?
<pxc>
and then vendors who want something stable and consistent to target can keep targeting it
<{^_^}>
[nixpkgs] @Synthetica9 opened pull request #37495 → mdv: init at 1.6.3 → https://git.io/vxnF0
blankhart has quit [Ping timeout: 256 seconds]
<MichaelRaskin>
Also, exploit writers
<pxc>
I know it's not BSD-like in the sense that with Nixpkgs NixOS & friends haven't taken ownership of the codebases that are packaged, like is done in the BSD world
<freeman42x>
llpp pdf reader doesn't seem to have any sufficient interface, I accidentally zoomed on some small bit of the screen and now it takes forever to zoom out
<pxc>
MichaelRaskin: definitely! it would have a big maintenance burden to avoid that. You know how RHEL has these dev-toolset repos that let you get newer versions of things installed under /opt?
<pxc>
I was picturing this as a less annoying way to achieve something similar
<duckwho>
freeman42x: you can map custom keybindings
<freeman42x>
zathura also is with minimal/no GUI
<pxc>
freeman42x: post some screenshots instead of just describing. :-P (and is your monitor high-dpi?)
<freeman42x>
duckwho: what I'm looking for is something like Foxit / Adobe
<freeman42x>
or Evince but that just works, without the blurriness issue
<duckwho>
freeman42x: Oh, then llpp wouldn't suit you...
<freeman42x>
pxc: I'd say blurry should pretty much describe it, but here I go :)
<infinisil>
MichaelRaskin: pbogdan: So this /bin/sh business, should I just ignore those failures too no hydra? Or can this fix (https://github.com/NixOS/nixpkgs/pull/36919) be ported to 18.03? (I still don't quiet get if that's even fixable and why it happens)
<MichaelRaskin>
pxc: re: maintenance: we don't do that good of a job doing something sane with stable
<MichaelRaskin>
Ideally the dependency on /bin/sh should be fixed in the builds, if only for the benefit of non-NixOS non-sandbox builds
<infinisil>
MichaelRaskin: By fixing you mean replacing it with ${pkgs.bash}/bin/sh ?
<pxc>
MichaelRaskin: I wasn't thinking of it as something for us to do. Just wondering about whether Nix could also be a useful tool to people who are interested in making more traditional enterprise-y operating systems, and what sort of impact it could make if they also perceived that usefulness and ran with it
<MichaelRaskin>
patchShebangs on something, something like that
<pxc>
and that kind of blurriness is not one I immediately recognize from the issues I've had :-(
fendor has quit [Quit: Leaving]
ottidmes has joined #nixos
blahdodo has quit [Ping timeout: 256 seconds]
justbeingglad has joined #nixos
<pxc>
freeman42x: I'd try Okular if you haven't, and also see whether the blurriness persists in other desktop environments or at other screen resolutions
blahdodo has joined #nixos
shachaf has quit [Ping timeout: 256 seconds]
<duckwho>
freeman42x: Oh btw. Does the mate version of evince have this problem too?
<pxc>
freeman42x: well, if it affects different DEs differently, probably they handle DPI scaling or font rendering differently. It gives you a place to look and compare
liori has quit [Quit: leaving]
dan_b has quit [Remote host closed the connection]
<freeman42x>
> freeman42x: well, if it affects different DEs differently, probably they handle DPI scaling or font rendering differently. It gives you a place to look and compare
<freeman42x>
well aware of that, but it doesn't matter anymore since I got one that works
<ottidmes>
does anyone has an example of sshfs automount in NixOS? I saw there was a autofs module that can do this for me, but I also read that you can do this with systemd, any examples of this?
hph^ has quit [Ping timeout: 246 seconds]
<s2mitrov>
what would you guys say is the most compelling reason to switch to nix over docker? i'm trying to shill nix to a friend, but the best i've got is better reproducibility due to more granular hash checking, nix being more lightweight + it doesn't always have to do a full rebuild if you change on small thing, and that you can use nix with docker for better control over the packages and stuff that get used to build your image
<Shados>
ottidmes: I'm using a systemd user service for it at home https://0x0.st/sB8p.txt, as per the bug linked in the desc you can't use .mount units for sshfs currently. Doesn't have the laziness of actual systemd automount, but handles failures and reconnects relatively OK, so eh.