<aszlig>
clever: anyway, for the deterministic partitioning issue... do you have any idea how to properly solve that?
<aszlig>
clever: i mean, right now i'm heading for a hybrid approach by encoding a SHA1 based on the devicespec into the UUID, but it obviously doesn't work for all file systems
<clever>
aszlig: either go entirely disk the label field, or find some PRNG that can generate UUID's
<clever>
aszlig: also, with gpt, you have 2 uuid's and 2 labels, one each on the partition table, and the filesystem
<aszlig>
clever: i can already generate the UUIDs for the file systems
<aszlig>
yeah, for gpt this is easy, but not on dos disklabels
seanz has quit [(Quit: Leaving.)]
<clever>
aszlig: i have also had no trouble getting legacy booting to work on gpt, so just say no to MBR :P
<aszlig>
my idea would be to punch a hole at the end of the partition layout and write the storage info into that
ryanartecona has quit [(Quit: ryanartecona)]
feepo has quit [(Quit: Connection closed for inactivity)]
<aszlig>
but having that in "unpartitioned area" also means that it might confuse users
stepcut has joined #nixos
<clever>
aszlig: you can always create a dedicated partition for it, thats what grub demands whenever you want legacy booting
<eacameron>
ok...very weird thing going on here. I have a function that takes some args and puts them into a URL for fetchzip. For some reason when I change the version string, nix is not detecting a change and thinks the closure is identical to the previous version. Other changes still trigger a rebuild, but not changing that string.
sitedyno has quit [(Ping timeout: 248 seconds)]
glines has joined #nixos
<eacameron>
ohh...weird. I forgot to change the sha. Changing that triggers rebuild.
<eacameron>
This might be a bug in fetchzip...
<NixOS_GitHub>
[nixpkgs] Mic92 closed pull request #16027: nixos/geoip-updater: new service (master...geoip-service) https://git.io/voTX2
<NixOS_GitHub>
[nixpkgs] Mic92 reopened pull request #16027: nixos/geoip-updater: new service (master...geoip-service) https://git.io/voTX2
AllanEsp1nosa has joined #nixos
sitedyno has joined #nixos
<gchristensen>
eacameron: nope
<gchristensen>
eacameron: which other parameters triggered the rebuild?
<gchristensen>
eacameron: the problem is that to nix, the only thing that matters is the sha. if it knows the sha it doesn't need to fetch it -- it already has it -- no matter if the URL changed
ilja_kuklic has quit [(Ping timeout: 252 seconds)]
<pikajude>
i thought that it *would* change if the derivation name changed?
<pikajude>
because both the hash and the name are in the path?
athan has quit [(Remote host closed the connection)]
orbekk has joined #nixos
obadz has joined #nixos
alx741_ has quit [(Quit: alx741_)]
alx741 has joined #nixos
<eacameron>
gchristensen: Hmm...but wouldn't *something* change? The derivation now has a new URL.
<eacameron>
gchristensen: I only checked version and sha. The only other parameter to that function was the name of the plugin, which was also part of the URL.
estewei has joined #nixos
<eacameron>
pikajude: That's an interesting point. I'm not including the version in the name of the package in this case.
<gchristensen>
pikajude: I'm not 100% sure, to be honest
<pikajude>
i know fetchurl puts the filename of the URL in the derivation name
<gchristensen>
but this really hurts users
<pikajude>
unless you change it
angerman has joined #nixos
georges-duperon has quit [(Ping timeout: 252 seconds)]
georges-duperon has joined #nixos
obadz has quit [(Read error: Connection reset by peer)]
derjohn_mob has quit [(Ping timeout: 256 seconds)]
anderslundstedt has quit [(Remote host closed the connection)]
<clever>
pikajude: i believe nix operates purely on the filename and hash, and if the filename hasnt changedm it can find the old tar, which still has the old hash
<clever>
pikajude: i always invalidate the hash by incrementing a random digit
anderslundstedt has joined #nixos
sdothum has quit [(Read error: Connection reset by peer)]
sdothum has joined #nixos
shanemikel has quit [(Quit: My MacBook Air has gone to sleep. ZZZzzz…)]
derjohn_mob has joined #nixos
bfrog has quit [(Quit: WeeChat 1.6)]
Lowl3v3l has quit [(Remote host closed the connection)]
<Profpatsch>
If I have DB data types and normal data types, where both are very similar but the DB types are a bit more serialized.
<Profpatsch>
How can I define isomorphisms for these in a clever way?
indi_ has joined #nixos
<Profpatsch>
Right now there’s functions like
<Profpatsch>
dbToTask (DBTask tit desc how url recur start) =
<Profpatsch>
Task tit desc how (importURL . toS <$> url) recur start
mbrgm has quit [(Ping timeout: 240 seconds)]
mbrgm has joined #nixos
saintromuald has joined #nixos
<Profpatsch>
pikajude: No, fetchurl uses fixed-input derivations.
<Profpatsch>
Those only use the sha
CodeWarrior has quit [(Ping timeout: 240 seconds)]
<Profpatsch>
Ah, I misread #nixos for #haskell again.
<Profpatsch>
sorry
<dtz>
hahaha I was wondering
<dtz>
like it wasn't /offtopic/ but I sure missed the context xD
<Profpatsch>
dtz: They are on #5 and #6 respectively in my weechat.
<Profpatsch>
So I mistype that sometimes.
<Profpatsch>
And often there is Haskell talk going on here or nix talk going on there, so that furter complicates matters. :P
<dtz>
haha exactly. NP, it happens :)
jsgrant_ has joined #nixos
jsgrant_ has quit [(Excess Flood)]
athan has joined #nixos
jsgrant_ has joined #nixos
hexagoxel has quit [(Ping timeout: 256 seconds)]
jsgrant_ has quit [(Ping timeout: 248 seconds)]
jsgrant_ has joined #nixos
stepcut_ has joined #nixos
<pikajude>
Profpatsch: really?
stepcut has quit [(Read error: Connection reset by peer)]
<pikajude>
That's bizarre
<pikajude>
I remember filenames being in the path
<pikajude>
And I've never seen two paths have the same hash but different name suffixes
hexagoxel has joined #nixos
deepfire has quit [(Ping timeout: 255 seconds)]
slack1256 has joined #nixos
mizu_no_oto has joined #nixos
jsgrant_ has quit [(Remote host closed the connection)]
jsgrant_ has joined #nixos
<slack1256>
how can I set up an alternative configuration for sshd on configuration.nix without definiting two different derivations?
<slack1256>
I basically need sshd to operate without password challenges but sometimes allow them to pass public keys
<slack1256>
yep managing to versions of configuration.nix to enable such setup is bothersome
mguentner has quit [(Quit: WeeChat 1.6)]
mguentner has joined #nixos
roconnor_ has joined #nixos
slack1256 has quit [(Remote host closed the connection)]
athan has quit [(Remote host closed the connection)]
marusich has joined #nixos
mkoenig has quit [(Ping timeout: 240 seconds)]
<NixOS_GitHub>
[nixpkgs] league opened pull request #21848: yuicompressor: add bin wrapper for jar (master...yuicompressor-bin) https://git.io/vM2Jy
mizu_no_oto has quit [(Quit: Computer has gone to sleep.)]
<eacameron>
This is more general than just Nix-related, but how would one go about making a script that sets some env vars and then drops the user into nix-shell. I want to provide a real, interactive shell; just with some changes to the env.
<clever>
eacameron: any attribute in a derivation becomes an env variable inside the nix-shell
mkoenig has joined #nixos
<clever>
eacameron: so you could just nix-shell -E 'with import <nixpkgs> {}; runCommand "dummy1" { buildInputs = [ hello ]; env1 = "valu1"; } "dummy2"'
<eacameron>
clever: no kidding. That's very cool. However, I'm not clear; I want to provide env vars *to* nix-shell
<clever>
which vars?
<eacameron>
clever: Actually what I want to do is have a script that sets NIX_PATH and then drops me into a shell.
<clever>
ah
<clever>
you can modify the nix-path with -I
<clever>
so just nix-shell -I nixpkgs=/home/clever/nixpkgs
<eacameron>
clever: Yah I realize that...but I was just ever so slightly hoping I could make it less manual than that. I actually grab that info from a config file relative to a repo root, etc. etc.
<clever>
ah
<eacameron>
clever: I'm trying to make a repo-based nix-shell essentially.
<eacameron>
Give me a shell where the nixpkgs is the one I picked for this project
<glines>
hrm... I'm trying to write a derivation for a .run file.... I'm trying how to make a "no op" for unpackCmd
<glines>
I don't have anything to unpack; I just want the .run file in the buildPhase
<clever>
eacameron: you could just make a 2 line bash script that does nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs-channels/archive/497e6d2f1d149f5fbe004e15fe8c384eed993943.zip
<eacameron>
clever: How's that? Wouldn't the bash script run and then return...leaving me in my original shell?
<clever>
eacameron: it will only return after nix-shell returns
<clever>
eacameron: so nix-shell will still give you a shell as it always does
danharaj has quit [(Ping timeout: 258 seconds)]
<glines>
clever: okay yeah. that plus a mkdir command works
<glines>
(nixpkgs expects a directory)
<clever>
glines: nixpkgs also uses a $sourceRoot to manage some things
<eacameron>
clever: wow...my intuition was completely broken on that one. You're right. I was thinking the subshell would break it.
<eacameron>
clever: Yah that works great!
Wizek_ has quit [(Ping timeout: 240 seconds)]
uralbash has joined #nixos
shanemikel has joined #nixos
georges-duperon has quit [(Ping timeout: 252 seconds)]
sdothum has quit [(Quit: ZNC - 1.6.0 - http://znc.in)]
dram_phone has joined #nixos
mguentner has quit [(Read error: Connection reset by peer)]
mguentner has joined #nixos
ryanartecona has joined #nixos
AllanEsp1nosa has quit [(Ping timeout: 256 seconds)]
systemfault has quit [(Quit: Bye!)]
deepfire has joined #nixos
glines has quit [(Ping timeout: 240 seconds)]
systemfault has joined #nixos
Sonarpulse has quit [(Ping timeout: 248 seconds)]
justanotheruser has quit [(Ping timeout: 256 seconds)]
justan0theruser has joined #nixos
c74d has joined #nixos
k0001 has quit [(Quit: Connection closed for inactivity)]
wkennington has joined #nixos
Havvy has joined #nixos
<shanemikel>
okay, trying to switch to git repo of nixpkgs.. so I added the repo to NIX_PATH as <git>, and aliased nix-git to nix-env -f "<git>".. and I created a new profile with `nix-env --switch-profile /nix/var/nix/profiles/git`.. then I switched back to the default profile to install nix and nss-cacert with `nix-git -p /nix/var/nix/profiles/git -i nix nss-cacert`
<shanemikel>
I'm having some problems with the binary cache and bad hashes
dram_phone has quit [(Read error: Connection reset by peer)]
simendsjo has quit [(Remote host closed the connection)]
ryanartecona has quit [(Quit: ryanartecona)]
dram_phone has joined #nixos
eacameron has quit [(Remote host closed the connection)]
abrar has quit [(Quit: WeeChat 1.6)]
eacameron has joined #nixos
dram_phone has quit [(Remote host closed the connection)]
athan has quit [(Ping timeout: 240 seconds)]
athan has joined #nixos
Geraldus has joined #nixos
MercurialAlchemi has joined #nixos
shanemikel has quit [(Quit: My MacBook Air has gone to sleep. ZZZzzz…)]
reinzelmann has joined #nixos
shanemikel has joined #nixos
Geraldus has quit [(Quit: Be back later ...)]
Geraldus has joined #nixos
<eacameron>
why is haskellPackages.time null?
Geraldus has quit [(Client Quit)]
<NixOS_GitHub>
[nixpkgs] bjornfor opened pull request #21850: nixos/prometheus: add configText option for alertmanager (master...prometheus) https://git.io/vM2mo
siarheit_ has joined #nixos
<simpson>
globin: Is bjornfor also working with you on Prometheus stuff? I have the blackbox_exporter working on a custom branch as well as the Python client bindings.
kampfschlaefer has joined #nixos
shanemikel has quit [(Quit: My MacBook Air has gone to sleep. ZZZzzz…)]
eacameron has quit [(Remote host closed the connection)]
marusich has quit [(Quit: Leaving)]
indi_ has quit [(Remote host closed the connection)]
indi_ has joined #nixos
eacameron has joined #nixos
yurrriq has joined #nixos
indi_ has quit [(Ping timeout: 240 seconds)]
<yurrriq>
Hey there. Is anyone in? I've got a package question.
eacameron has quit [(Ping timeout: 240 seconds)]
mkoenig has quit [(Ping timeout: 240 seconds)]
<yurrriq>
Generally, there's an existing package and I want to derive another that has one added dependency and changes three or four attributes from the old package.
<yurrriq>
I've got a working version that's mostly copy and paste, but that feels wrong.
<yurrriq>
I'm familiar with overrideDerivation, but I'm not sure how to add another dependency there..
mrkgnao has joined #nixos
mkoenig has joined #nixos
<yurrriq>
Specifically, I want to take the existing lilypond-2.18.2 (stable) package and write an unstable one, namely 2.19.24 (for now)
eacameron has joined #nixos
smola_ has quit [(Ping timeout: 255 seconds)]
Bodil has quit [(Ping timeout: 255 seconds)]
BramD has quit [(Ping timeout: 255 seconds)]
codedmart has quit [(Ping timeout: 255 seconds)]
bret has quit [(Ping timeout: 255 seconds)]
alphor has quit [(Ping timeout: 255 seconds)]
jonasfj has quit [(Ping timeout: 255 seconds)]
jonasfj has joined #nixos
Bodil has joined #nixos
smola_ has joined #nixos
hexa- has quit [(Ping timeout: 255 seconds)]
puffnfresh has quit [(Ping timeout: 255 seconds)]
carter has quit [(Ping timeout: 255 seconds)]
mpickering has quit [(Ping timeout: 255 seconds)]
fresheyeball has quit [(Ping timeout: 255 seconds)]
nz has quit [(Ping timeout: 255 seconds)]
johnsonav has quit [(Ping timeout: 255 seconds)]
thoughtpolice has quit [(Ping timeout: 255 seconds)]
aristid has quit [(Ping timeout: 255 seconds)]
frankpf has quit [(Ping timeout: 255 seconds)]
jgrasser has quit [(Ping timeout: 255 seconds)]
cyraxjoe has quit [(Ping timeout: 255 seconds)]
echo_oddly has quit [(Ping timeout: 255 seconds)]
ericbmerritt_ has quit [(Ping timeout: 255 seconds)]
mbrock has quit [(Ping timeout: 255 seconds)]
Forkk has quit [(Ping timeout: 255 seconds)]
mjacob has quit [(Ping timeout: 255 seconds)]
fresheyeball has joined #nixos
cyraxjoe has joined #nixos
jgrasser has joined #nixos
mjacob has joined #nixos
codedmart has joined #nixos
Forkk has joined #nixos
eacameron has quit [(Ping timeout: 245 seconds)]
thoughtpolice has joined #nixos
aristid has joined #nixos
johnsonav has joined #nixos
carter has joined #nixos
echo_oddly has joined #nixos
mpickering has joined #nixos
nz has joined #nixos
katyucha_ has joined #nixos
<dbe>
For the "Move the Wiki!" milestone - the wiki is moving to nixpkgs/nixos/doc/manual, right?
mbrock has joined #nixos
ericbmerritt_ has joined #nixos
frankpf has joined #nixos
alphor has joined #nixos
puffnfresh has joined #nixos
wkennington has quit [(Read error: Connection reset by peer)]
hexa- has joined #nixos
BramD has joined #nixos
katyucha has quit [(Killed (verne.freenode.net (Nickname regained by services)))]
katyucha_ is now known as katyucha
systemfault has quit [(Quit: Bye!)]
katyucha_ has joined #nixos
<dbe>
I guess there is also nix/doc/manual and nixpkgs/doc (apparently not /doc/manual :P). But I guess that's where it's going. I'll try my luck at one of these tickets.
<Myrl-saki>
Well...
<Myrl-saki>
Is it only me, or is nix-prefetch-url currently incompatible with steam URLs?
<Baughn>
So, yes, it'll take a while to build. I recommend installing Chromium in user context.
<the-kenny>
Yeah, that's the issue here. Hydra doesn't built it and therefore it isn't available in the binary cache
<the-kenny>
Actually, I just use google-chrome instead of chromium on my machine
<ixxie>
Baughn: how does using it in user context help?
<Baughn>
It'd avoid needing to rebuild chromium to change your machine config.
<Baughn>
Or use google-chrome instead, yeah.
<ixxie>
hmmm....well I cannot go that route because I am working on a completely declarative config to share with my team
<Baughn>
If you're enabling the plugins anyway, you might as well.
<ixxie>
yeah
<ixxie>
I guess thats what I wouild do
derjohn_mob has joined #nixos
ThatDocsLady has joined #nixos
ThatDocsLady has quit [(Read error: Connection reset by peer)]
roconnor__ has joined #nixos
Ralith_ has quit [(Ping timeout: 240 seconds)]
<NixOS_GitHub>
[nix] edolstra pushed 1 new commit to 1.11-maintenance: https://git.io/vM24Q
<NixOS_GitHub>
nix/1.11-maintenance 1fa2c86 Eelco Dolstra: Bump version to 1.11.6
roconnor_ has quit [(Ping timeout: 258 seconds)]
<NixOS_GitHub>
[nixpkgs] lheckemann opened pull request #21852: syncthing: remove runtime dependency on go (master...remove-syncthing-go-dep) https://git.io/vM24p
ilja_kuklic has quit [(Ping timeout: 240 seconds)]
athan has joined #nixos
athan has quit [(Ping timeout: 240 seconds)]
thetet has joined #nixos
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
roconnor__ has quit [(Quit: Konversation terminated!)]
Itkovian has joined #nixos
<ixxie>
Baughn: for some reason I cannot find the google-chrome package anywhere (not on nix-env -qa nor on the website package search)
<dramforever>
I can install it as a nix package right?
<ixxie>
LnL: is there a query function more compatible with system wide declarative configuration?
<LnL>
ixxie: not that I know of, the best think I can think of is nixpkgs.config = import /root/.nixpkgs/config.nix;
<LnL>
but that only works for a single user
kuznero has joined #nixos
<kuznero>
Hi All!
<ixxie>
alright LnL, I will make an alias ^^
Geraldus has quit [(Quit: Be back later ...)]
<LnL>
dramforever: yes every package from hackage in available in nixpkgs :)
<ixxie>
LnL: way- you mean putting that in the nixos user configuration
<dramforever>
LnL: but the build products are gone?
<kuznero>
Can anybody recommend a way to solve kind of problems when `/nix/store/wfpv2d26vmqqpqirqins5i702yp19j0n-fsharp-4.0.1.1/lib/mono/4.5/FSharp.Core.dll` path does not exist any longer (probably after one of the nix-collect applied)? Should I be able to simply reinstall packages under question (fsharp and mono46 in this case)?
<LnL>
ixxie: no that's for you configuration.nix, make it use your user config so both are the same
teknico has joined #nixos
<LnL>
ixxie: you can put a config.nix file next to configuration.nix and symlink it to all the users or something
<ixxie>
LnL: this seems to be another symptom of the lack of user configuration infrastructure in NixOS
Minoru has joined #nixos
<LnL>
yes, I have some ideas about that but I'm busy with other things again
<ixxie>
aye
<Minoru>
hi. I am migrating from Funtoo to NixOS and I have trouble with my ~/.XCompose file: https://raw.githubusercontent.com/Minoru/dotfiles/master/.XCompose The include doesn't work because NixOS doesn't create /usr/share at all. I ran `find / -name Compose` and it seems like the only place where Compose files are referenced is /nix/store, i.e. Nix doesn't symlink to them from anywhere. (I did manually check
<Minoru>
~/.nix-profile and /run/current-system/, just to be sure. No symlinks.) Any ideas on what else I can try?
<LnL>
ixxie: but if it's just for files you can use the environment.etc module
kuznero has quit [(Ping timeout: 260 seconds)]
<ixxie>
LnL: all I wanna do is have a command to query all packages xD
rexim has joined #nixos
<ixxie>
LnL: so I really don't understand what your last comment is about (and I am kinda curious what you think I was trying to do)
<LnL>
ixxie: well, if you want that you can create an alias in your configuration
kuznero has joined #nixos
<mrkgnao>
how do I use the GitHub version of a library?
<mrkgnao>
the expression in nixpkgs uses a specific (fairly old) commit, and I want to use the latest
<kuznero>
Sorry, got disconnected. Does anybody have any experience in working with F# on mono in NixOS? I started getting errors like this after I ran `nix-collect`: Microsoft.Common.targets: error : Cannot copy /nix/store/wfpv2d26vmqqpqirqins5i702yp19j0n-fsharp-4.0.1.1/lib/mono/4.5/FSharp.Core.dll to /.../bin/Debug/FSharp.Core.dll, as the source file doesn't exist.
civodul has quit [(Quit: ERC (IRC client for Emacs 25.1.1))]
<kuznero>
Or perhaps this is more generic problem when one package is made aware of one version of another package, but after upgrade the second package had changed its position... something like that maybe?
MercurialAlchemi has quit [(Ping timeout: 256 seconds)]
<ixxie>
LnL: I am still not clear on what that alias should look like
<rexim>
Minoru: sup
MercurialAlchemi has joined #nixos
<avn>
Ohh, I see folks work on some .net/mono stuff ;) And I have a question then. How proper deal with `nuget restore` stuff in build process, when packaging stuff to nixpkgs?
<mrkgnao>
can anyone help me get the GitHub master version of a package?
<LnL>
ixxie: then I think you can run nix-search \* or nix-search -A haskellPackages
sigmundv has quit [(Disconnected by services)]
<kuznero>
avn: I am not sure as I am not creating nix packages with mono projects. I am using NixOS as my development OS and just want to make mono+F# work together again
<mrkgnao>
I'm guessing I have to change the args of fetchFromGithub
<ixxie>
LnL: lets find out :) what is this \* ? lists everything?
<mrkgnao>
seems like an escaped regex
<LnL>
everything toplevel, it doesn't look inside attributes (like the haskellPackages set)
<mrkgnao>
-ish thing
<mrkgnao>
LnL: how do you look inside attribute?
sigmundv_ has joined #nixos
<mrkgnao>
attributes*
<LnL>
like other example, with -A
<kuznero>
How one would inspect what is it in user setup (what nix packages installed by user) and what by system-wide setup?
ixxie has quit [(Quit: Lost terminal)]
<srhb>
kuznero: For the user, nix-env -q
<kuznero>
ok, it is simply `nix-env -q` :) and then I will try to remove fsharp and mono from system-wide and user-specific, call collect in both modes again and after try to re-install fsharp and mono and hope it will work together after that
<kuznero>
srhb: thanks
ixxie has joined #nixos
mkoenig has quit [(Ping timeout: 255 seconds)]
<kuznero>
nope, that simply didn't work. But what I noticed is that after I remove fsharp and mono from system and user and called collect it didn't remove anything from the system. Why is that so? It kept all the related files in the system no matter....
mkoenig has joined #nixos
rexim has quit [(Remote host closed the connection)]
<gchristensen>
kuznero: it stays in the /nix/store until it is garbage collected
<avn>
kuznero: my kid want to play minecraft, and I think about package truecraft (which is C#/gtk#)
<unlmtd[m]>
is there a nix clone with copyfree license?
<gchristensen>
unlmtd[m]: is there a specific license you're looking for?
<unlmtd[m]>
copyfree
iMatejC has quit [(Remote host closed the connection)]
<unlmtd[m]>
otherwise said, a non-license license. a license that doesnt enforce anything, (what I would call free)
<gchristensen>
I don't believe so
kuznero has quit [(Ping timeout: 260 seconds)]
<unlmtd[m]>
theres some haskell work which is permissive
<gchristensen>
srhb: yeah I didn't know how to compress my concern in to just a few words, sorry :)
<gchristensen>
niksnut: what version of hydra does hydra.nixos.org run? the footer just says abcdef
<mrkgnao>
srhb: so what you meant is basically use channels/16.09 and merge the ... patches ... that updated the package you want to the latest version
<LnL>
^ yeah I noticed that :p
<mrkgnao>
I think the correct term is "cherry-pick"?
<ixxie>
hmmm LnL looks like there is no --arg / --option flag on nix-env --query
<srhb>
mrkgnao: Don't create the patches on top of stable though
<srhb>
mrkgnao: Make them on unstable, then cherry onto your stable branch :)
<srhb>
mrkgnao: Stable merging is iffy due to security patches (I discovered last weekend)
<srhb>
So if you ever want to PR, it's much easier from an unstable or master patch. :)
<srhb>
mrkgnao: So my workflow is: checkout nixos-unstable. Create changes here, then cherry-pick them onto my stable checkout for regular use. :)
<LnL>
ixxie: why did you think --arg/--argstr doesn't work with nix-env
<dtz>
haha, reminds me of http://bash.org/?152037 (not that you were trolling just "It can't be done" -> summons help immediately like nothing else haha)
eacameron has joined #nixos
blgria has joined #nixos
Fare has joined #nixos
<ixxie>
LnL: because the query didn't succeed, but I guess more because I misunderstood the manual
<gchristensen>
niksnut: hrm... are you running that based on 16.09 or based on unstable or something else? running that on unstable has provided run-time errors about bzip2, and running that on stable has provided assorted compile-time errors
<ixxie>
LnL: that did the trick! thanks..... I wasn
<LnL>
gchristensen: huh, I just switched and it looks fine
<ixxie>
I wasn't pipping to grep (I still didn't get the search argument syntax, and I guess most people don't because many grep for stuff)
<gchristensen>
LnL: wat?
<gchristensen>
LnL: can you share your config?
<LnL>
ixxie: nix-env -qaP '.*chrome.*' ...
<NixOS_GitHub>
[nixpkgs] globin pushed 1 new commit to master: https://git.io/vM21x
<NixOS_GitHub>
nixpkgs/master e2be232 Robin Gloster: perlPackages.CompressRawBzip2: 2.064 -> 2.070
<gchristensen>
well look at that, globin
<gchristensen>
like magic I think globin fixed it ;)
civodul has joined #nixos
<globin>
%)
<globin>
didn't even read that here
<niksnut>
gchristensen: on 16.09, but with a patch to nixUnstable
<kuznero>
just collecting garbage and removing generations does not seem to be enough....
<MoreTea>
@kuznero, you mean that you want to reset /etc/nixos/configuration.nix to what the installer created?
<MoreTea>
With *.nix files you describe how your system should look like. You should version those files yourself if you want to have different generations of them. Nix is not a VCS.
<MoreTea>
Does that answer your question?
FareTower has joined #nixos
mog has joined #nixos
<kuznero>
MoreTea: not exactly... currently I ran into the problem that I have 2 packages (mono46 and fsharp). fsharp is dependent on mono (not mono46) package and thus nix installs both mono and mono46. After I figured this out I made a package override for fsharp to depend on mono46 instead of mono. I rebuilt it all, but on a new system it helps, whether on the current old nixos it does not.
<kuznero>
MoreTea: so, I am trying to figure out how to undo all the changes that have been done to the system other than in /etc/nixos/**/*.nix files
mudri has joined #nixos
<sphalerite>
by rebuilding
<MoreTea>
nixos-rebuild switch
<MoreTea>
your entire NixOS "version" is upgraded by the nixos-rebuild tool
<LnL>
yes, just revert the changes in /etc/nixos and rebuild
<kuznero>
This was the first thing I have tried
<sphalerite>
if you want to delete the old stuff because it's using up space, delete the old generations of the system configurations
<gchristensen>
niksnut: can this merge? is there a problem with it? V
<kuznero>
Does not really help with that specific fsharp/mono46/mono issue
<sphalerite>
what is the issue?
<LnL>
if you don't have the old configuration anymore you can rollback, but then you won't be able to move forward anymore untill you fix your configuration
<kuznero>
sphalerite: currently I ran into the problem that I have 2 packages (mono46 and fsharp). fsharp is dependent on mono (not mono46) package and thus nix installs both mono and mono46. After I figured this out I made a package override for fsharp to depend on mono46 instead of mono. I rebuilt it all, but on a new system it helps, whether on the current old nixos it does not.
<gchristensen>
it sounds like maybe running fsharp wrote a config file or something which persisted what mono it is calling?
<kuznero>
How about I remove all mono/mono46/fsharp and run `collect-grabage -d` + delete generations + nix-store --optimise?
<sphalerite>
kuznero: what do you mean by it helping?
<kier>
is there an easy way to check which nixpkgs commit the nixpkgs-unstable channel is currently equivalent to?
<gchristensen>
kuznero: you don't need to optimize, just delete all refs to mono and collect garbage
<sphalerite>
you want mono to be gone?
<kuznero>
gchristensen: that might be a very sound idea. then it is the matter of cleaning my user profile...
<goibhniu>
kier: if you follow the URL for the channel you'll see the hash at the end
<kuznero>
sphalerite: no, I need fsharp to be referring to mono46 and not mono
<sphalerite>
It should already be referring to mono46 if you made the override and rebuilt.
sdothum has joined #nixos
<kuznero>
Yes, but it is not
<kuznero>
On a new machine from this config it does reference mono46, but on my it is still mono
<sphalerite>
and you installed it where? in systemPackages or in your user environment (using nix-env)?
<kuznero>
systemPackages
<gchristensen>
kuznero: you may also consider using strace to see what _exactly_ is going on
<kuznero>
gchristensen: is strace just a package that I can install and use?
<gchristensen>
sure, `nix-shell -p strace` but you'll probably want to look up how to use it
johann__ has joined #nixos
<kuznero>
gchristensen: thanks, will do
<gchristensen>
good luck :) it sounds like either you've gotten to be pretty good with nix, doing overrides but have missed something simple, or fsharp is doing something wacko :)
<kuznero>
gchristensen: google to the rescue :)
MoreTea has quit [(Remote host closed the connection)]
mrkgnao has quit [(Ping timeout: 240 seconds)]
aminechikhaoui has quit [(Ping timeout: 255 seconds)]
<gchristensen>
:*
<gchristensen>
:)
cpennington has joined #nixos
<kuznero>
gchristensen: Thanks! That was exactly that... I simply removed all of the ~/.mono ~/.nuget ~/.config/.mono and ~/.config/xbuild, and it just works after that. So, it was simply caching
<gchristensen>
kuznero: annoying! if you dig in more and find out what is going, maybe it can be "fixed" / patched in nixpkgs to not behave so poorly
MoreTea has joined #nixos
<kuznero>
gchristensen: it feels like it needs to be fixed in xbuild/mono infrastructure
<kuznero>
to retry without assuming data from local cache
johann__ has quit [(Read error: No route to host)]
thetet has joined #nixos
<ixxie>
I am getting warning: name collision in input Nix expressions, skipping ‘/home/ixxie/.nix-defexpr/channels_root/nixos’
<ixxie>
with nix-env commands
guillaum1 has joined #nixos
<guillaum1>
Hi today. There is something which stills bug me. How can I ensure reproducible build when my nix files does not specify any version number?
<gchristensen>
how do you mean?
<MoreTea>
Ah, I figured out my issues with minikube (_and_ nixops); I accidentally installed virtualbox from the unstable channel
<MoreTea>
Reverting back to the one in 16.09 made things work.
<MoreTea>
(because i'm running the vbox module provided in 16.09)
<garbas>
MoreTea: ixxie: my idea only works for nix-build/nix-shell/nix-env command, for nixos-rebuild you need to set NIX_PATH=nixpkgs=https://..
johann__ has joined #nixos
<garbas>
there use to be an option in nixos to pin nixpkgs and module system would re-evaluate itself
<garbas>
but it was removed since it broke some other stuff :)
<MoreTea>
I installed virtualbox via nix-env -i :x
<MoreTea>
I probably should use systemPackages for those programs.
<ixxie>
garbas: I guess for most purposes, NixOS stable is 'reproducible enough'.... I would go with your approach for publishing scientific research though, and it that scenario it would be a default.nix anyway
kuznero has quit [(Quit: Page closed)]
<gchristensen>
I'm not sure what level of reproducibility guillaum1 is referring to
<gchristensen>
bit-for-bit vs. "same version of software"
Sonarpulse has joined #nixos
<guillaum1>
gchristensen: I'm talking about getting a build against the same .so and .h files.
<gchristensen>
bit-for-bit identical?
<gchristensen>
or same input source and compiler
<guillaum1>
gchristensen: nop, I really don't care about bit-for-bit. Just I know that I tested with version X of a package and I don't want it to be replaced by version Y
<gchristensen>
well if your version of nixpkgs stays the same it'll be the same
<gchristensen>
when you build stuff with nix you have to declare all the inputs it has, and if it needs network access you have to tell nix ahead of time what the sha of the result will be
<guillaum1>
gchristensen: for the dependency I'm handling myself, yes. But for, say, openexr, I'm happy with the package inside nixpkg, but I want to ensure that it will still be the same in a few months
<guillaum1>
ixxie: thank you, I'm reading it.
eacameron has joined #nixos
<gchristensen>
guillaum1: as I said, you would have to keep your version of nixpkgs stable
georges-duperon has joined #nixos
<guillaum1>
gchristensen: if by stable you mean keeping the same git clone hash, that's alright.
<gchristensen>
right
<gchristensen>
but you should indeed try new versions over time as important security updates get applied
eacameron has quit [(Remote host closed the connection)]
<guillaum1>
gchristensen: I totally agree with that. But the current status is that someone installed an ubuntu 8 years ago and we are using it to build everything and we are praying for it to work. So I considere pinning a version of nix as a great improvment ;)
<gchristensen>
yes
<guillaum1>
That's really great ;)
eacameron has joined #nixos
<sphalerite>
why is the attribute called linuxHeaders26Cross when it's actually linux 4.4 headers?
mudri has quit [(Read error: Connection reset by peer)]
mudri has joined #nixos
onla has joined #nixos
eacameron has quit [(Read error: Connection reset by peer)]
eacameron has joined #nixos
<ixxie>
sphalerite: I get linuxHeaders for 2.2 headers
<ixxie>
4.4*
<sphalerite>
ixxie: that's not Cross though
<sphalerite>
also, I installed firefox ESR (nix-env -iA nixos.firefox-esr) and nix-env -u wants to upgrade it to the non-ESR version. Can I prevent that from happening while still getting upgrades when the ESR gets upgraded?
<gchristensen>
out of curiousity -- why do you like esr? (not saying you shouldn't use it, just never met a user before)
eacameron has quit [(Ping timeout: 245 seconds)]
lsix has quit [(Ping timeout: 256 seconds)]
lsix has joined #nixos
mizu_no_oto has quit [(Quit: Computer has gone to sleep.)]
<ixxie>
sphalerite: oops :P
<sphalerite>
fewer updates = less space used on nixos :p
<gchristensen>
sphalerite: are you in a space constrained environment?
<sphalerite>
yes, my root fs is 25GB
<sphalerite>
I was previously on debian where it was never a problem but I often find myself battling the constraints on nixos
<sphalerite>
other than that, no major reason to use the ESR.
<sphalerite>
Citing @SwiftOnSecurity? I approve :D
michas has joined #nixos
<sphalerite>
Right, I think it's time I fixed my space issues once and for all
<sphalerite>
(it's only a matter of partitioning)
reinzelmann has quit [(Ping timeout: 258 seconds)]
thetet has quit [(Quit: Leaving.)]
<Sonarpulse>
shlevy: where are the release notes?
<Sonarpulse>
release-notes.xml seems embarrisingly old
eacameron has joined #nixos
<ixxie>
speaking of space, just discovered nix.gc.automatic - pretty sweet!
<gchristensen>
niksnut: can you set up a webhook on the nixpkgs repository to send hooks to a project of mine? polling github 86,400 times a day makes them very unhappy. I don't have the URL setup yet, but it will be https://webhook.nix.gsc.io/nixos/nixpkgs
<Sonarpulse>
^^^^^ +9999
<gchristensen>
Sonarpulse: hmmm? :)
<Sonarpulse>
I approve of this request :)
eacameron has quit [(Read error: Connection reset by peer)]
<shlevy>
Sonarpulse: Probably best to put it in the nixos release notes, unfortunately :(
<Sonarpulse>
shlevy: yeah
<Sonarpulse>
was just going to say
<Sonarpulse>
I found that
Rotaerk has quit [(Read error: Connection reset by peer)]
<Sonarpulse>
and saw strictly nixpkgs things in there
<shlevy>
Sonarpulse: Check with globin though, the new release manager
<shlevy>
Maybe we should bring pack separate release notes
hlavaty has quit [(Read error: Connection reset by peer)]
<Sonarpulse>
shlevy: yeah, I wouldn't be opposed to retroactively splitting out nixpkgs changes, etc
<Sonarpulse>
might just do nixos for now, and leave that future work
altphi has joined #nixos
<ixxie>
does it make sense to run auto gc for a laptop? what happens if its off at the designated time
<shlevy>
ixxie: I think that run is just skipped
<shlevy>
ixxie: IIRC it's implemented with systemd timers
cpennington has quit [(Ping timeout: 240 seconds)]
jmeredith has joined #nixos
<Sonarpulse>
shlevy: one reason for my rush is trying to figure out my vs DavidEGrayson's cross PR
<shlevy>
Right
<Sonarpulse>
I think mine is better long term, but his had/has tests passing
<ixxie>
thanks shlevy
<Sonarpulse>
I'm thinking now, even if I'd end up reverting some of his abstractions
<Sonarpulse>
good to get that foundation of tests passing
<Sonarpulse>
so "normalize...", his, my "sane cross.." would be ideal merge order
<shlevy>
If we're just going to end up reverting his changes, can't you just bring the tests into yours?
hlavaty has joined #nixos
<Sonarpulse>
shlevy: I did do some of that already
<Sonarpulse>
top-level/release-cross.nix has been beefed up
kampfschlaefer has joined #nixos
<Sonarpulse>
but I still change a few hashes for unclear reasons
<Sonarpulse>
break a few other things
<Sonarpulse>
but I was working from a state that's already very weird
<Sonarpulse>
most packages are grossly confused on host vs target (in the autoconf sense)
johann__ has quit [(Quit: Leaving.)]
<Sonarpulse>
which makes getting things working without fixing all the packages a bit of a delicate balancing act
FareTower has quit [(Ping timeout: 240 seconds)]
<shlevy>
I don't think no-hash-changing is a fundamental requirement
<shlevy>
It's nice, of course, but we have hydra
<Sonarpulse>
shlevy: I agree
sellout- has quit [(Quit: Leaving.)]
<Sonarpulse>
but I like to know *why* the hash changes
<Sonarpulse>
if it's not intentional, so far its been a good indicator of bugs
<shlevy>
Ah, yeah
<Sonarpulse>
I found a way to half-do my cross change
<Sonarpulse>
such that most cross hashes didn't change
<Sonarpulse>
but some random ones did, including the big cross bootstrap tools jobs, which were broken
<Sonarpulse>
travis doesn't doing release-cross.nix too, which complicates things
<Sonarpulse>
perhaps I should push a new branch to nixpkgs and make a hydra job for it?
<Sonarpulse>
I've found ways to exact use from them
<Sonarpulse>
but it is a pain
<Sonarpulse>
neeed to rebase nixos-unstable, right after it changes
alx741_ has joined #nixos
alx741 has quit [(Ping timeout: 260 seconds)]
<gchristensen>
(perhaps I should bump ikwildrpepper / domenkozar on this one) can you set up a webhook on the nixpkgs repository to send hooks to a project of mine? polling github 86,400 times a day makes them very unhappy. I don't have the URL setup yet, but it will be https://webhook.nix.gsc.io/nixos/nixpkgs
<qknight>
ltspool, a binary, uses libunbound.so which i can see in NIX_LDFLAGS but the rpath to the library is not added at all so later the binary won't work since the library can't be found. now i wonder how NIX_LDFLAGS works in the first place since i've added buildPhase = '' ... something'' so maybe it is not used at all?!
cpennington has joined #nixos
<qknight>
any idea welcome
michas has quit [(Ping timeout: 260 seconds)]
eacameron has joined #nixos
altphi has quit [(Quit: altphi)]
asseadinho has joined #nixos
erasmas has joined #nixos
reinzelmann has joined #nixos
eacameron has quit [(Remote host closed the connection)]
eacameron has joined #nixos
fritschy has quit [(Remote host closed the connection)]
<asseadinho>
Hi all, made a newbie mistake and would appreciate any help.
<asseadinho>
I can't `nixops deploy` anymore, after I set `services.openssh.permitRootLogin = "no";`
<asseadinho>
I still have a shell open, but can't figure out how to re-enable ssh so that I can re-depoy.
<asseadinho>
Any ideas?
<shlevy>
asseadinho: Root shell?
<asseadinho>
yep
<shlevy>
asseadinho: I don't remember the nixops command line exactly, but there should be one that lets you just build the system derivation
<shlevy>
asseadinho: Then you can nix-copy-closure from the shell and run the 'switch-to-configuration' script
MercurialAlchemi has quit [(Ping timeout: 256 seconds)]
FareTower has joined #nixos
<asseadinho>
ok, I'm going to google all that. thanks!
angerman has quit [(Quit: Gone)]
FareTower has quit [(Ping timeout: 245 seconds)]
michas has joined #nixos
mizu_no_oto has joined #nixos
ThatDocsLady has quit [(Ping timeout: 252 seconds)]
katyucha has quit [(Quit: leaving)]
<qknight>
another question: with nix-shell i can either use -A foo or -p vim but not both at the same time?
thc202 has quit [(Ping timeout: 252 seconds)]
justan0theruser has quit [(Ping timeout: 240 seconds)]
phanimahesh has joined #nixos
glines has joined #nixos
pie_ has joined #nixos
michas has quit [(Ping timeout: 256 seconds)]
<ToxicFrog>
Is there any way to get nixos-rebuild (boot|switch) to name the boot entries with a datetime rather than just an index?
<sphalerite>
resize successful! I now have 72 beautiful gigabytes free :D
<ToxicFrog>
Or better yet, annotate it with arbitrary text?
ixxie has quit [(Ping timeout: 255 seconds)]
<pie_>
nix should be written in javascript, the it would be easy to change </trolling>
* pie_
runs
<glines>
Ralith: Did you ever get a vulkan development environment going?
<glines>
I'm just trying to figure out the "proper" way to set up a vulkan app. The sdk is pre-compiled binaries iinm
<offlinehacker[m]>
Where is served manual for nixos-unstable, all links on Hydra seems to have disappered?
<offlinehacker[m]>
Also, why are links on Hydra missing?
<glines>
offlinehacker[m]: hrm. IDK about hydra, but you could enable services.nixosManual.showManual = true;
srhb has quit [(Quit: Quit)]
<asseadinho>
shlevy: `nixos-rebuild switch --rollback` was sufficient. I'm rookie indeed.
mkoenig has quit [(Ping timeout: 248 seconds)]
<glines>
it shows up in the F8 terminal with that option
reinzelmann has quit [(Ping timeout: 255 seconds)]
srhb has joined #nixos
estewei has quit [(Quit: Quit)]
mkoenig has joined #nixos
<glines>
oh
<glines>
Ralith_: Did you ever get vulkan development going?
<offlinehacker[m]>
glines: this thing is not really helpful, I was asking if we serve somewhere older docs, good luck copying something from tty8
<offlinehacker[m]>
Newer docs*
<gchristensen>
I'm not aware of a place hosting docs for unstable
<offlinehacker[m]>
gchristensen: it was on Hydra, but now all links are missing
<gchristensen>
ah, likely since moving the build products to s3
<offlinehacker[m]>
gchristensen: s3 supports serving files over HTTP, and as I know we do that, looks like Hydra does not have updated links
<gchristensen>
yes it is after all an http addressable object store
<glines>
If you have nixpkgs checked out locally, you could do nix-build nixos/release.nix -A manual.x86_64-linux
<sphalerite>
I have https://gist.github.com/lheckemann/8c0968598b777d894765a3dadcd77383 which isn't building right because the toolchain is EABI while the default configuration for uclibc is OABI. I believe all I need to do is add CONFIG_AEABI=y to the extraConfig for the uclibc expression, but how would I do that in this case?
<NixOS_GitHub>
[nixpkgs] globin pushed 2 new commits to master: https://git.io/vMak3
<LnL>
the haskell community is generally just to strict about their version bounds for some reason
<shlevy>
It makes sense
<shlevy>
they use a modified semver
<shlevy>
You shouldn't bump the upper bound above a possibly breaking change without testing it, as an upstream
<simpson>
It's too bad that Haskell packages don't come with tests more often. Then we could rely on the test results to indicate whether a package is broken.
<shlevy>
Yeah
* shlevy
mumbles something about dependent types
<simpson>
It's also too bad that Haskellers don't use monover. Far easier than semver.
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<glines>
With nixpkgs, how does CMake manage to find header directories? looking at the output of env, I see my header directories show up in NIX_CFLAGS_COMPILE... but surely cmake doesn't know how to use that?
<domenkozar>
shlevy: yeah too late now :)
simendsjo has joined #nixos
<glines>
there's no such thing as INCLUDE_PATH...
<shlevy>
Hmm interesting, I thought 'network' automatically did ipv6Only = false
<simpson>
shlevy: Well, yes, if your language fails to bake in certain security guarantees, then trying to chase after them later is futile.
<shlevy>
domenkozar: Bah. Just need good enough type inference :P
<domenkozar>
shlevy: well really that code is just wrong
<simpson>
But that can always be handled by your distro, which can expand monover to semver and use the patch level to indicate distro-specific patches.
<domenkozar>
we should've binded one socket with ipv6only = false
<domenkozar>
but it's hard to debug these errors
ertes has quit [(Ping timeout: 240 seconds)]
<domenkozar>
since HasCallStack doesn't help
<Sonarpulse>
simpson: I do hate the sloppy definition of semver
<domenkozar>
or I'm incompetent to use it properly
<Sonarpulse>
but third digit for non-api changing is interesting
<Sonarpulse>
compatability is not a symmetric relation
asseadinho has quit [(Ping timeout: 240 seconds)]
<shlevy>
I thought you didn't need withSocketsDo if you're using mkSocket
<simpson>
Sonarpulse: It isn't? I mean, it seems to be once you stop trying to differentiate between forwards- and backwards-incompatibility.
<Sonarpulse>
but the 3rd digit indicates a equivalence subrelation within the compatability relation
<Sonarpulse>
simpson: oh but you must!
<shlevy>
Sonarpulse: modulo bugs
<domenkozar>
shlevy: it's no harm done
<Sonarpulse>
shlevy: yeah, not modulo bugs there'd be way fewer compatabilites :)
<domenkozar>
well who's going to write apidiff
<Sonarpulse>
simpson: the 2nd digit (and monover's 1st?) indicates compatability in one direction
<simpson>
Sonarpulse: Why? I only care about whether my current collection of packages all work together; if I want to choose a different version of a package that is still compatible, I must try each version in turn, plugging it into the rest of my package set and running all of the compatibility tests.
<shlevy>
This is why we should have fully specified interfaces, down to performance guarantees, machine checked :P
nschoe has quit [(Quit: Program. Terminated.)]
<Sonarpulse>
I love the package version problem
<Sonarpulse>
because all the cool stuff helps
<simpson>
shlevy: Sure, but Haskell's not gonna be that language. Not with its current module system.
<Sonarpulse>
machine checked stuff computing compatability relation: proof theory
<Sonarpulse>
trust and distributed builds: crypto
<shlevy>
simpson: I'm hopeful for C++'25 to be the one!
<Sonarpulse>
solving all this: model theory
<simpson>
shlevy: I'm sticking with Monte. Tired of types.
<Sonarpulse>
all the verification techniques each have something to simultanously contribute! :)
<Sonarpulse>
simpson: I think haskell could do it except for template haskell
<Sonarpulse>
elm does, I hear?
<shlevy>
does typed TH help?
endformationage has joined #nixos
<Sonarpulse>
probably
<Sonarpulse>
welll
<Sonarpulse>
:)
Shou has quit [(Ping timeout: 240 seconds)]
<Sonarpulse>
you can observe TH term I think
mckeankylej has joined #nixos
<Sonarpulse>
even typed one
<Sonarpulse>
so lololol not quite
<shlevy>
Oh
<shlevy>
Well of course assuming the TH terminates
<mckeankylej>
what is the name of the nix function that writes a script to the nix store? Its like writeScriptBin or something
<Sonarpulse>
I mean like case (asdf :: Th (a -> B)] { Lambda _ _ _ -> ; ... }
<simpson>
Sonarpulse, shlevy: Here are two problems that I don't think Haskell has good answers for. First, is it possible to encapsulate a value s.t. you can statically prove that the value is not accessed outside of a certain scope?
<Sonarpulse>
simpson: backpack
pi3r has joined #nixos
<Sonarpulse>
even if you are meh on backpack's new features
<Sonarpulse>
the background work answers these sorts of questinos
<Sonarpulse>
ezyang will save us!
<simpson>
Sonarpulse: No, you're right, Backpack is a *real* module system, and Haskell has needed it.
<Sonarpulse>
yeah, fancy features or not, "packages" have semantics in the language and that should be formalized, not just fall out of cabal implementation ad-hocly
<simpson>
Second, is it possible to run Haskell code at runtime to prove facts about values? Can you use it to e.g. prove that there's no mutable values reachable from a given value?
<Sonarpulse>
do you need to do that?
<simpson>
Of course not. Neither of these are need-to-have features. AFAICT perfect encapsulation is frightningly rare, but nobody seems to mind.
<Sonarpulse>
I doubt that is possible, but I'm not sure what you are trying to achieve
<Sonarpulse>
the type of encapsulation that's needed for this I think is entirely static
<simpson>
Well, in Monte, if I say `x :DeepFrozen`, I've asserted that x is an immutable object; it contains no mutable references. This is the *bottom* of Monte's auditor runtime.
<simpson>
This kind of encapsulation is statically scoped, sure.
<simpson>
And it provides *cap-safety*, wherein I can make assertions about how my program will not leak data in certain ways.
<Sonarpulse>
was monte the one with the normal forms with debrauin?
<pi3r>
does someone know if there is issue with github nixpkgs archive ?
<pi3r>
I was doing -I nixpkgs=https://github.com/NixOS/nixpkgs-channels/archive/9d0a9bc9d61f6e94d720486ddbe972af8547382d.tar.gz but it doesn't seem to work anymore
<Sonarpulse>
will do
guillaum1 has joined #nixos
eacamero_ has joined #nixos
<simpson>
Sonarpulse: But basically DeepFrozen and other auditors are allowed to see the object's AST and binding guards, which specify the structure and types of values within the object.
<Sonarpulse>
ok interesting
<Sonarpulse>
shlevy: btw about to push again
<simpson>
And it's possible to infer a *wealth* of object behaviors from that. We build immutability, equality, etc. all the way up to pass-by-copy semantics with these.
<Sonarpulse>
cool!
<Sonarpulse>
that is a very prinicipled approach for dynamic types
ryanartecona has joined #nixos
eacameron has quit [(Ping timeout: 252 seconds)]
<simpson>
I *really* need to write that blog post about static types and The Network.
<Sonarpulse>
which I didn't realize until now was made by same author
<Sonarpulse>
probably explains both doing the normal forms good for networking
<simpson>
Sonarpulse: Morte and Dhall are fun ideas. They have the problem that they are very foundational; I do *not* want to work in anything remotely as low-level as Morte.
<simpson>
Morte's dependency injection ends up being exactly the kind of factoring that leads to safe eval().
<simpson>
And safe eval() is like *the* cool thing about Monte.
edvorg has joined #nixos
<Sonarpulse>
simpson: internesting hmmm
guillaum1 has quit [(Ping timeout: 240 seconds)]
deepfire has joined #nixos
lsix has quit [(Quit: Konversation terminated!)]
Shou has quit [(Ping timeout: 245 seconds)]
derjohn_mob has quit [(Ping timeout: 240 seconds)]
<Ralith>
glines: yeah, working well now
pi3r has quit [(Ping timeout: 240 seconds)]
justin_ has joined #nixos
<glines>
Ralith: I figured it out, mostly
<glines>
I didn't notice the vulkan-loader.dev derivation
<glines>
right now I'm trying to figure out cmake...
<justin_>
hello, I'm working on a simple webservice for practicing deployment in nixos. it is build in a default.nix (nix-build default.nix produces the desired output) and the corresponding service with its dependencies is defined in a file called bierboerse.nix. Can somebody give me a tip, how I reference my binary from default.nix best in bierboerse.nix (the service definition)?
<glines>
It finds the vulkan library and even prints out the path, but ${VULKAN_LIBRARY} is still an empty string in cmake somehow <_<
<glines>
I'm doing something completely stupid
<glines>
justin_: in the nix language, you can reference your derivation with ${myderivation} inside any string
<glines>
justin_: the ${myderivation} syntax gets expanded to the path to the output of your derivation
<glines>
something like ${myderivation} ---> /nix/store/149f29...
<glines>
justin_: to get a string to your binary, you would do ${myderivation}/bin/mybinary
<glines>
so you would get ${myderivation} ---> /nix/store/149f29.../bin/mybinary
altphi has joined #nixos
<glines>
justin_: now with the path to your binary, you can use ln to make a symbolic link to it. or you can stuff it in a config file. or... various other uses
deepfire has quit [(Ping timeout: 256 seconds)]
mudri has quit [(Ping timeout: 248 seconds)]
<Sonarpulse>
shlevy: should be good to go!
<Sonarpulse>
if you are still around :fingers-crossed:
<regnat[m]>
Sonarpulse: typed nix-like langauge :), no way to do our big fix-extends probably :( <- you might be interested by https://www.gofundme.com/typing-nix :)
eacamero_ has quit [(Remote host closed the connection)]
<regnat[m]>
(honnesty note : I'm the one behind this)
__Sander__ has quit [(Quit: Konversation terminated!)]
<Sonarpulse>
regnat[m]: will take a look!
<Ralith>
glines: are you using current nixos-unstable?
<Ralith>
glines: also note that on nix projects you don't really need cmake scripts like that, since all your deps end up in sane search paths
MoreTea has quit [(Ping timeout: 240 seconds)]
<Sonarpulse>
regnat[m]: woah!
michas has joined #nixos
<pierron>
regnat[m]: There is no way to make NixOS in a typed manner either.
<regnat[m]>
pierron: What do you mean by that ?
guillaum1 has joined #nixos
<pierron>
regnat[m]: I mean that option definitions are quite versatille, and checking that each option has the right type would be hard, within the module system.
<Ralith>
pierron: nixos could use refinement and/or gradual types just fine and still benefit
<regnat[m]>
Oh yes, It would be hard
edvorg has quit [(Ping timeout: 240 seconds)]
<pierron>
regnat[m]: the function which recursively process the module system has to dynamically check for the type based on the options which are providing the location of each name.
<pierron>
Ralith: I honestly never headr of gradual types before this one.
<regnat[m]>
But there are some great possibilities still
<Ralith>
glines: the working vulkan infrastructure isn't present in 16.09; your build environment needs to be on unstable to compile correctly, and your OS needs to be on unstable (or at least cherry-pick some key commits) to run it
<Sonarpulse>
IMO, its better to accept some refactoring as part of the project
<Sonarpulse>
than try to make it fit *exactly*
yenzenz has quit [(Ping timeout: 240 seconds)]
<Ralith>
pierron: well, I encourage you to study the prior art on retrofitting type systems to dynamic languages
<regnat[m]>
And static typing isn't incompatible with dynamic testing
<Ralith>
especially before you dismiss it as "there is no way"
<Ralith>
e.g. Typed Scheme is quite pleasant
<pierron>
Ralith: I do not dismiss it, I just said that NixOS option definitions cannot be given a type ahead of time, otherwise you would be duplicating the logic which is already within the module system.
<regnat[m]>
Yes, typed scheme is a quite relevant type system
danharaj has joined #nixos
<regnat[m]>
(btw pierron, if you're the pierron who designed the module system and are still in Paris, i'd be very happy to discuss with you sometime)
<pierron>
regnat[m]: yes, and ok.
<regnat[m]>
They still can be given some typing constraints
<NixOS_GitHub>
[nixpkgs] jpierre03 opened pull request #21861: update write-mirror-list.sh with idea from issue #8561 (master...patch-2) https://git.io/vManG
<Sonarpulse>
if/case refinements can help make the dynamic checks work for nix
glines has quit [(Ping timeout: 258 seconds)]
<Sonarpulse>
*for module system
<regnat[m]>
Which can be very helpful, given that the modular design of the module system makes it hard to test
eacameron has joined #nixos
pi3r has joined #nixos
hlavaty has quit [(Remote host closed the connection)]
systemfault has joined #nixos
<NixOS_GitHub>
[nixpkgs] 7c6f434c pushed 2 new commits to master: https://git.io/vManS
<NixOS_GitHub>
nixpkgs/master 4e0a5e7 Peter Hoeg: nim: include all supporting tools...
<NixOS_GitHub>
nixpkgs/master 94b0ad1 Michael Raskin: Merge pull request #21846 from peterhoeg/f/nim...
eacameron has quit [(Read error: Connection reset by peer)]
<Ralith>
glines: also I don't think anyone's tested/patched the nvidia drivers' vulkan support yet, though they might Just Work
georges-duperon has quit [(Ping timeout: 252 seconds)]
<justin_>
@glines thank you, so I need to add the default.nix to my imports?
Avaq has quit [(Ping timeout: 245 seconds)]
<justin_>
glines: thank you, so I need to add the default.nix to my imports?
<shlevy>
Sonarpulse: reviewing
<NixOS_GitHub>
[nixpkgs] 7c6f434c pushed 1 new commit to master: https://git.io/vMaCV
<NixOS_GitHub>
nixpkgs/master bc2e81f Michael Raskin: i3blocks: fix platforms
eacameron has joined #nixos
ryanartecona has quit [(Quit: ryanartecona)]
eacameron has quit [(Ping timeout: 240 seconds)]
<NixOS_GitHub>
[nixpkgs] vcunat pushed 1 new commit to master: https://git.io/vMaWZ
<NixOS_GitHub>
nixpkgs/master a652099 Vladimír Čunát: fetchurl: change to grep -E to fix #8561...
<NixOS_GitHub>
[nixpkgs] vcunat closed pull request #21861: update write-mirror-list.sh with idea from issue #8561 (master...patch-2) https://git.io/vManG
<Sonarpulse>
shlevy: thanks!!!
ryanartecona has joined #nixos
cpennington has quit [(Ping timeout: 240 seconds)]
michas has quit [(Ping timeout: 252 seconds)]
AllanEspinosa has quit [(Ping timeout: 248 seconds)]
fresheyeball has quit [(Ping timeout: 258 seconds)]
georges-duperon has joined #nixos
eacameron has joined #nixos
fresheyeball has joined #nixos
phanimahesh has quit [(Ping timeout: 255 seconds)]
Itkovian has joined #nixos
eacameron has quit [(Ping timeout: 240 seconds)]
mudri has joined #nixos
c0bw3b has joined #nixos
pi3r has quit [(Ping timeout: 255 seconds)]
cpennington has joined #nixos
eacameron has joined #nixos
phanimahesh has joined #nixos
vozz has joined #nixos
pie_ has quit [(Ping timeout: 240 seconds)]
eacameron has quit [(Ping timeout: 255 seconds)]
<vozz>
Anyone got any idea why chromecast wouldn't be working from chromium? I've disabled my firewall which I thought might be a problem, but it still says "No Cast destinations found. Need help?"... I can stream from lots of other devices, so it isn't a problem with the network or the chromecast itself.
<gchristensen>
I think there is some udev shenanigans ?
<vozz>
what do you mean by shenanigans? :P
* jack[m]
is watching #21268.
<jack[m]>
I'm a bit curious as to the path to being able to cross-compile a rpi image, though.
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<Dezgeg>
full nixos is probably quite unfeasible as things like getting perl to cross compile is tough
<Dezgeg>
but maybe something minimal with only kernel, busybox, nix and the like where you can build the rest locally would be feasible
<Sonarpulse>
jack[m]: oh, thanks for watching!
<Sonarpulse>
I am Ericson2314 :)
guillaum1 has quit [(Ping timeout: 255 seconds)]
guillaum1 has joined #nixos
<avn>
Dezgeg: kernel requires perl to build ;)
<avn>
Dezgeg: but as I know, some work in progress done to remove runtime dependencies on perl from nix
<Dezgeg>
afaik kernel doesn't, or at least I've seen "die perl die"-style posts to lkml
<Dezgeg>
and it would be a native perl anyway
<regnat[m]>
avn : But does it require perl on the target system ?
eacameron has joined #nixos
<avn>
regnat[m]: atm yes, idk how 0.12.x branch is ready to use
<avn>
pkgs/build-support/buildenv/builder.pl also need some work ;)
athan has quit [(Ping timeout: 240 seconds)]
<Dezgeg>
but buildenv would only need a native perl, not cross perl
<NixOS_GitHub>
[nixpkgs] Ericson2314 pushed 9 new commits to master: https://git.io/vMa0o
<NixOS_GitHub>
nixpkgs/master 0ef8b69 John Ericson: top-level: Modernize stdenv.overrides giving it self and super...
<NixOS_GitHub>
nixpkgs/master 3e197f7 John Ericson: top-level: Normalize stdenv booting...
<NixOS_GitHub>
nixpkgs/master 0f33b9f David Grayson: top-level: Do stdenvOverrides in stage.nix even if crossSystem exists....
vozz has quit [(Ping timeout: 240 seconds)]
athan has joined #nixos
eacameron has quit [(Ping timeout: 255 seconds)]
glines has joined #nixos
<glines>
Ralith: I am on the latest nixos-unstable
<glines>
and I'm running the Vulkan demos....
<Ralith>
glines: if you're in a nix-shell with vulkan-loader, then /nix/store/zym29l7hvp4b2n45qhhh2k8dd98aj2sr-vulkan-loader-1.0.26.0-dev/include should be in your include path and everything should Just Work
<glines>
agreed xD
<Ralith>
it's not clear to me why you're even using this cmake script
vozz has joined #nixos
<Ralith>
just pass -lvulkan and #include <vulkan/vulkan.h>
<glines>
eh in principle I will ask try to build on Windows
<glines>
also
guillaum1 has quit [(Quit: Lost terminal)]
<glines>
eh I figured it out. it wasn't even using that script
eacameron has joined #nixos
georges-duperon has quit [(Ping timeout: 252 seconds)]
stkj has joined #nixos
abbradar has joined #nixos
<DavidEGrayson>
Sonarpulse: If you're around, I'm getting less excited about doing cross-compiling within nixpkgs. I think I might just write my own recipes for making cross-toolchains and cross-compiling the software I care about, which would be outside of nixpkgs and just use nixpkgs as a dependency. It might be a lot of work, but it would give me more control and the result would be easier to understand
<DavidEGrayson>
(nixpkgs has so many *if* statements all over the place when compiling GCC).
<abbradar>
Hi all! Is there a way to find out Nix store path Nix was compiled with? Need that for Debian packaging improvements
<DavidEGrayson>
Sonarpulse: Oh, I see you pushed some stuff to nixpkgs/master just a few minutes ago, that's exciting I guess.
<DavidEGrayson>
Sonarpulse: Oh, you have commit access now.
stkj has left #nixos []
<DavidEGrayson>
abbradar: Maybe Nix is already in the nix store so you can just see where it is? I'd try $(which nix) and $(readlink $(which nix))
<abbradar>
DavidEGrayson: it won't be on Debian and other non-NixOS distributions
<abbradar>
we install it to /usr/bin as usual
<abbradar>
(I'm patching Nix now to add `nix-store --print-store-dir`)
<NixOS_GitHub>
[nixpkgs] primeos opened pull request #21862: [RFC] Refactor the firewall (master...firewall) https://git.io/vMagl
<Sonarpulse>
DavidEGrayson: I am around
<Sonarpulse>
DavidEGrayson: yeah I got commit last winter/fall I forget
<Sonarpulse>
DavidEGrayson: I'd encorage you not to give up hope!
<Sonarpulse>
DavidEGrayson: honestly take a look at my "sane .." PR
<Sonarpulse>
especially in a few moments when I rebase it
<Sonarpulse>
I ****remove**** a top of those conditionals
<Sonarpulse>
I suspect long term the gcc derivations can be dirastically simplified too
<Sonarpulse>
though I am waiting to do that
<Sonarpulse>
if you could help me test the sane PR
<Sonarpulse>
I think we'd get something in good shape quite quickly
<DavidEGrayson>
Sonarpulse: :) Cool, maybe I will. Not this weekend but maybe in a while. So yeah, I don't really feel penalized if my pull request sits around a while longer. I probably shouldn't have pinged people to look at it.
shanemikel has joined #nixos
<DavidEGrayson>
Sonarpulse: I made a recipe for cross-binutils last night, and there was a lot of junk I could remove (as compared to the nixpkgs binutils) because I am not building the binutils with bootstrap tools.
k0001 has joined #nixos
<abbradar>
Sonarpulse: you mentioned that "sane..." PRs are yours -- are you Ericson2314?
<Sonarpulse>
abbradar: with sane cross, instead of hacking up mkDerivation, I hack up callPackage
<abbradar>
Sonarpulse: (still there with you)
<Sonarpulse>
and have multiple package sets (as normall bootstrapping also has multiple package sets)
<gchristensen>
is there a way to erase everything about a declarative nixos containter and start it fresh?
<abbradar>
gchristensen: I'd go destroy and create
<Sonarpulse>
soo callPackakge does the { nativeDrv = ...; crossDrv = ...; }
<k0001>
When I `nix-shell` into some Haskell project, the bash shell I'm dropped into is a bit flawed: The arrow up doesn't work and the prompt looks like "\[\][nix-shell:foo]$\[\]" for example. What's wrong?
<abbradar>
not sure if there's a better way
<Sonarpulse>
splicing the packaget ses together
<gchristensen>
abbradar: does that work with declarative containers?
<DavidEGrayson>
k0001: I'm not sure, I've used nix-shell on Linux with Haskell projects just fine.
<Sonarpulse>
then, I deprecate (they are still there) "system" "platform" and "crossSystem"
ixxie has joined #nixos
<k0001>
When I press the arrow up, I get something like "^[[A" echoed to the console. Similar for other keys such as "delete".
<Sonarpulse>
and do "localSystem = { inherit system platform; }" "hostSystem" "targetSystem"
<Sonarpulse>
DavidEGrayson: ^ that list bit helps avoid conditional madness
<abbradar>
Sonarpulse: so we have two "hidden" derivations created by callPackage, with mkDerivation choosing right ones when expanding to `derivation`?
<abbradar>
k0001: that happened to me when I ran nix-shell inside nix-shell
<abbradar>
gchristensen: nope, for declarative ones I think I've rm -rf'ed something
<gchristensen>
abbradar: like stop / rm -rf /var/lib/containers/... / start?
<abbradar>
Sonarpulse: not sure how "platform" and "system" differs now; what's "platform"?
<Sonarpulse>
abbradar: um that's kind of a mix up of old and new :)
<abbradar>
gchristensen: yep
<Sonarpulse>
abbradar: system is a string, platform is ....other stuff
<abbradar>
oh... now I remember
<abbradar>
I've actually encountered platform
<k0001>
abbradar: ahh! That was it. Thanks!
<Sonarpulse>
crossSystem.platform is a thing
Itkovian has joined #nixos
<Sonarpulse>
that is the same basically
<Sonarpulse>
crossSystem.config is *almost* system, except is proper LLVM target tripple
<gchristensen>
abbradar: neat, thank you
<Sonarpulse>
well need a to make localPlatform.config and {build,host}Platform.system
<Sonarpulse>
but w/e future work
<Sonarpulse>
abbradar: as to your first bit, before mkDerivation is secretely a wrapper making these hidden derivations (and for nativeBuildInputs etc will use the right ones)
<Sonarpulse>
now mkDerivation *just* selects the right ones for buildInputs nativeBuildInputs etc
<abbradar>
Sonarpulse: I concluded my messed up understanding from the thought that _something_ still needs to differ between nativeBuildInputs and buildInputs
<abbradar>
Oh yes, then I understood correctly
<abbradar>
mkDerivation just chooses right ones; callPackage actually makes native and cross versions
<abbradar>
correct?
ryanartecona has quit [(Quit: ryanartecona)]
<Sonarpulse>
closer :)
<Sonarpulse>
plain old bootstrapping
<Sonarpulse>
makes the right versions
<Sonarpulse>
callPackage is enlisted to splice together the boostrap stages to fake the old api
<Sonarpulse>
for mkDerivation to pick the right ones
<Sonarpulse>
one awkwardness here, is if you *override* callPackage's filter arguemnts, you'll need to do the picking yourself
mizu_no_oto has joined #nixos
<Sonarpulse>
I bring up that awkwardness cause it helps explain the division of labor
<abbradar>
you basically say "let the host (I can be wrong about terminology here, need to refresh host vs build) be stage-1, target = stage", correct?
sellout-1 has quit [(Read error: Connection reset by peer)]
sellout-2 has joined #nixos
sellout- has quit [(Ping timeout: 255 seconds)]
cpennington has quit [(Ping timeout: 258 seconds)]
justin_ has joined #nixos
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
Itkovian has joined #nixos
ckauhaus has quit [(Quit: Leaving.)]
vagrant- has joined #nixos
mog has joined #nixos
<vagrant->
hey guys, does anyone have experience running apache as built by nix? I'm trying to run it via systemd and it's hanging on startup. I think, though I'm not sure, that this is because it's trying to write to a folder in the nix repo. Has anyone experience this, and/or knows how to get a healthy apache running?
<simpson>
vagrant-: Known issue. I think that it's been fixed but I'm not sure. copumpkin was working on it, I think?
<NixOS_GitHub>
[nixpkgs] pSub pushed 2 new commits to master: https://git.io/vMaMJ
<NixOS_GitHub>
nixpkgs/master ea7a8bf Kier Davis: ckb: init at 0.2.6...
<vagrant->
simpson: it looks like nixos has some machinery to build a systemd unit for httpd, which might work? is it broken on nixos, or just other OSes? I'm trying to run it on Centos7
<vagrant->
if the nixos one works, is there some easy way to build it (or just build the canned httpd.conf) without copy/pasting code? I'm not sure if the nixpkgs top-level package set exposes any of those objects, or if I can perhaps get it from something in the nixos directory?
<simpson>
vagrant-: Oh, not NixOS. No idea, sorry. It works on NixOS, except for the write-to-root bug.
ixxie has quit [(Ping timeout: 252 seconds)]
<vagrant->
I don't really know much about the nixos package structure. I'm looking at nixos/modules/web-servers/apache-httpd/default.nix, and it's meant to be passed a 'config' argument. Is there any easy way to build the 'confFile' derivation in there?
cpennington has joined #nixos
endformationage has joined #nixos
reinzelmann has quit [(Quit: Leaving)]
<simpson>
vagrant-: No, things in nixos/ are usually not usable outside of NixOS.
mudri has joined #nixos
ntc has joined #nixos
vozz has quit [(Ping timeout: 252 seconds)]
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
Jackneill has joined #nixos
<vagrant->
hmm, that's a bit of a shame. Well, I suppose I can copy/paste something
Itkovian has joined #nixos
<simpson>
Patches welcome, but this just isn't what NixOS was designed for.
<NixOS_GitHub>
[nixpkgs] pSub pushed 2 new commits to master: https://git.io/vMaym
<vagrant->
yeah that's fair. However there's nothing stopping a working systemd unit or config file from being ported to another system running systemd, as long as its dependencies live in the nix store... at least, AFAIK
<simpson>
I guess? It's not something that's ever worked for me.
simendsjo has quit [(Remote host closed the connection)]
eacameron has quit [(Quit: Leaving...)]
<pmade>
I'd like to deploy a website via NixOps but if the site is in /nix/store Apache will always report Last-Modified as 1/1/1970. Any ideas/workarounds?
<simpson>
pmade: Don't serve from the store.
<pmade>
simpson: Simple answer. I was thinking I could copy them out of the store but then I'd have two copies of them.
<pmade>
I guess rsync is always an option.
<simpson>
pmade: And they'd still have those last-modified times, because *everything* in the store has that mtime.
<pmade>
Does NixOps have a way to deploy things that aren't in the store?
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
altphi has quit [(Quit: altphi)]
mkoenig has joined #nixos
mrkgnao has quit [(Ping timeout: 260 seconds)]
zarel has joined #nixos
cyraxjoe has quit [(Ping timeout: 256 seconds)]
cyraxjoe has joined #nixos
altphi has joined #nixos
sternenseemann has joined #nixos
<sternenseemann>
hey guys, is there a convenient way to, either a) override the platform with a custom-defined one or b) override the kernelExtraConfig?
catch22 has quit [(Quit: Leaving)]
xadi has quit [(Ping timeout: 258 seconds)]
mizu_no_oto has quit [(Quit: ["Textual IRC Client: www.textualapp.com"])]
aminechikhaoui has quit [(Ping timeout: 258 seconds)]
hiratara has quit [(Ping timeout: 255 seconds)]
cpennington has quit [(Remote host closed the connection)]
mkoenig has quit [(Ping timeout: 260 seconds)]
justin_ has joined #nixos
phanimahesh has quit [(Ping timeout: 240 seconds)]
mkoenig has joined #nixos
hiratara has joined #nixos
mkoenig has quit [(Read error: Connection reset by peer)]
aminechikhaoui has joined #nixos
bennofs1 has joined #nixos
<cheshircat>
hey, I'm trying to use nixops with a local libvirtd on nixos, and even though I have deployment.headless = true, it tries to connect to a graphical console and gives me an error
<cheshircat>
Is the nixops manual out of date, or is there a problem with nixos-unstable, or am I missing something?