<NixOS_GitHub>
[nixpkgs] NeQuissimus pushed 1 new commit to release-17.03: https://git.io/vQ5NF
<NixOS_GitHub>
nixpkgs/release-17.03 c82b517 Tim Steinbach: perf: Apply patch for offline kernels...
NixOS_GitHub has left #nixos []
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] NeQuissimus pushed 1 new commit to master: https://git.io/vQ5NN
<NixOS_GitHub>
nixpkgs/master 954c669 Tim Steinbach: perf: Apply patch for offline kernels...
NixOS_GitHub has left #nixos []
takle has quit [(Remote host closed the connection)]
newhoggy has quit [(Ping timeout: 268 seconds)]
newhoggy has joined #nixos
mexisme has joined #nixos
newhoggy has quit [(Ping timeout: 240 seconds)]
zeus_ has quit [(Remote host closed the connection)]
fetchurl has quit [(Ping timeout: 260 seconds)]
zeus_ has joined #nixos
mexisme2 has quit [(Ping timeout: 240 seconds)]
zeus_ has quit [(Ping timeout: 246 seconds)]
newhoggy has joined #nixos
newhoggy has quit [(Ping timeout: 248 seconds)]
vaibhavsagar has joined #nixos
zeus_ has joined #nixos
newhoggy has joined #nixos
newhoggy has quit [(Ping timeout: 248 seconds)]
spinus has quit [(Remote host closed the connection)]
newhoggy has joined #nixos
newhoggy has quit [(Ping timeout: 240 seconds)]
vaibhavsagar has quit [(Remote host closed the connection)]
vaibhavsagar has joined #nixos
newhoggy has joined #nixos
<gchristensen>
quiet friday evening
eacameron has quit [(Remote host closed the connection)]
hellrazo1 has joined #nixos
newhoggy has quit [(Ping timeout: 260 seconds)]
mexisme has quit [(Ping timeout: 260 seconds)]
Wizek_ has quit [(Ping timeout: 255 seconds)]
hellrazor has quit [(Ping timeout: 255 seconds)]
taktoa has joined #nixos
dbmikus has joined #nixos
newhoggy has joined #nixos
<Infinisil>
indeed
<gchristensen>
who wants to write some docs?
<Infinisil>
for what?
<gchristensen>
anything in nixos you've ever felt needs more docs :)
<Infinisil>
Ah, yes that would be good, or contribute to the unofficial wiki
newhoggy has quit [(Ping timeout: 260 seconds)]
<gchristensen>
sure
<gchristensen>
I'm trying to make the type docs in the module system clearer
<gchristensen>
https://nixos.org/nixos/manual/#idm140737315337648 is very strange "Value types are type that take a value parameter. The only value type in the library is enum." <goes on to list three>
<Infinisil>
Heh
<gchristensen>
I don't even know howto fix that :P
<Infinisil>
gchristensen: Have you ever read through the full manual?
<gchristensen>
no, some parts of it are a bit too painful. have you?
eacameron has joined #nixos
<gchristensen>
also a str isn't a basic type?
<Infinisil>
Nah, just searching/skimming through them
<Infinisil>
string is a basic type *and* a string related type, that's probably why it's like this
<Infinisil>
But I agree, should be in the basic type category
<gchristensen>
yeah
<gchristensen>
I think there needs to be a headline for the mergable string types, but I'm not sure what that would be
newhoggy has joined #nixos
thc202 has quit [(Ping timeout: 276 seconds)]
<gchristensen>
also I'm not sure Multiple definitions are concatenated with a new line "\n". is pedantically correct (though it doesn't really matter) but I think saying "Multiple definitions are joined by new lines (\n)" would be more clear
<Infinisil>
A sequential format for such things isn't optimal imo
<gchristensen>
oh?
eacameron has quit [(Ping timeout: 276 seconds)]
<Infinisil>
In general, text is rarely linear throughout, e.g. I often take notes and group them in chapters and such, but there are connections inbetween all of them
<gchristensen>
yeah
eacameron has joined #nixos
<gchristensen>
I'm just not sure how else to do it. what do you think?
<Infinisil>
Not sure
<gchristensen>
:)
<manveru>
i'd like to have links :)
<Infinisil>
Doesn't readthedocs have a search bar?
<gchristensen>
manveru: to what, from where?
<Infinisil>
yeah links would be nice, links to the source would even be nicer
<manveru>
i'm a bit tired of having to look up functions in nixpkgs... there's no comprehensive documentation of them either
<Infinisil>
gchristensen: Yes
<gchristensen>
to be sure I'm not interested in debating it tonight, I'm working on them tonight :)
nslqqq has quit [(Ping timeout: 240 seconds)]
<gchristensen>
I found a small section I'd like to make better
<manveru>
it's 3:30am or so here :| not gonna start on docs now
newhoggy has quit [(Ping timeout: 276 seconds)]
<gchristensen>
sure
<Infinisil>
Well before I even wanna start discussing docstrings again, I'll first try to build something that would do that and then come back and show it to get opinions
<manveru>
a lot of methods in lib have comments, just parsing that would be a nice thing already
<manveru>
s/methods/functions
<gchristensen>
there are alternatives that don't require parsing fwiw
<gchristensen>
fancy parsing, anyway
nslqqq has joined #nixos
Sonarpulse has quit [(Ping timeout: 260 seconds)]
eacameron has quit [(Ping timeout: 260 seconds)]
<Infinisil>
Like?
<gchristensen>
I guess I'll stop improving docs and see if I can get creative
<manveru>
also i feel a lot of the build-support stuff could be moved to lib
<Infinisil>
Well I know one, but I don't think it's what you think of
<gchristensen>
oh, Infinisil?
<Infinisil>
I'll try to keep the explanation short
newhoggy has joined #nixos
lambdamu_ has joined #nixos
hellrazo1 has quit [(Ping timeout: 248 seconds)]
<Infinisil>
A "doc shell" kinda thing, where you could interactively add, edit and view docs to everything, this would be stored in a database (not in source), and could tell you if docs are outdated or not, could generate manual from it, and maybe more things
<gchristensen>
that could be cool for sure
<Infinisil>
You could even make stuff such as "nix doc maintain" which would choose a random function that has outdated docs/no docs, shows it to you and asks you to document it
lambdamu has quit [(Ping timeout: 260 seconds)]
<gchristensen>
hey,that sounds good
newhoggy has quit [(Ping timeout: 260 seconds)]
newhoggy has joined #nixos
<Infinisil>
Another thing this would allow is to edit docs in whatever format you want, md/xml/html, which gets all converted to the same thing under the hood, but it could present you the format of your choice
<gchristensen>
Infinisil: do you want to hack on this idea? you can play if you want
<gchristensen>
if not, I might explore
<Infinisil>
Go ahead, it's too late for me to be productive anyways
newhoggy has quit [(Remote host closed the connection)]
<gchristensen>
ok but if I make this you have to write lots of doc strings :P
taktoa has quit [(Remote host closed the connection)]
newhoggy has joined #nixos
<Infinisil>
I might!
s33se has joined #nixos
eacameron has joined #nixos
mbrgm has quit [(Read error: Connection reset by peer)]
<gchristensen>
builtins.unsafeGetAttrPos will be helpful here
Supersonic112 has quit [(Disconnected by services)]
Supersonic112_ has joined #nixos
Supersonic112_ is now known as Supersonic112
<Infinisil>
For when a test is failing?
<gchristensen>
no, for linking to where the function is defined :)
s33se_ has quit [(Ping timeout: 276 seconds)]
<Infinisil>
Ahh yes
<Infinisil>
If I would build such an (experimental) doc shell, do you think it matters what language I'd use? Because I have no clue about C++, but all of nix uses it..
<gchristensen>
I have no idea
mbrgm has joined #nixos
eacameron has quit [(Ping timeout: 260 seconds)]
<TimePath>
I wouldn't think it matters too much, if it's an optional utility and not required for building derivations
<TimePath>
tried cross compiling something that depended on perl to arm once, didn't go well
newhoggy has quit [(Remote host closed the connection)]
eacameron has joined #nixos
<Infinisil>
Well I'd just use whatever language I want
<Infinisil>
If people would ever want it to be integrated and feel the need for a rewrite in C++ they could do that
cpennington has joined #nixos
<simpson>
Build the first draft, *then* worry about choice of language.
<Infinisil>
And integrating a different language wouldn't be a problem anyways
<Infinisil>
simpson: I'd just start hacking something together first
cpennington has quit [(Remote host closed the connection)]
<gchristensen>
"NOTE: this function is not performant and should never be used."
* gchristensen
uses it
<Infinisil>
Damn, why does it even exist when it shouldn't be used
<TimePath>
I googled that
<TimePath>
first result: "5 Phrases You Should Never Use In An Email"
eacameron has quit [(Read error: Connection reset by peer)]
<simpson>
gchristensen: I'm convinced but I don't have commit bit.
<gchristensen>
that's fine, you have experience with nix
newhoggy has joined #nixos
<Infinisil>
Man, why do I have so many (seemingly great) ideas but am always too lazy to do anything. I always feel like I should be studying instead (I'll get expelled if I don't), so I just contemplate all day
<Infinisil>
And end up doing nothing
<dash>
Infinisil: write them down
<simpson>
Infinisil: Don't worry, most of your ideas, like most ideas in general, aren't worth pursuing. (No idea which ideas you're referring to right now.)
<simpson>
Like dash says, write them down. Spend one sentence of ten words, and two sentences of twenty words, per idea that you like.
<Infinisil>
dash: Doing that
<simpson>
*Then*, set them down, wait a day, and see if any of them are still good.
<gchristensen>
or don't do them and don't worry about it, whatever, its fine
newhoggy has quit [(Ping timeout: 240 seconds)]
<simpson>
Oh yeah, never forget, there's three outcomes of sitting around smoking a pipe and saying "Wouldn't it be cool if...?"
<Infinisil>
This sounds sketchy as hell, but these ideas I'm having seem to solve so many problems in todays software
<simpson>
(1) "I've solved the murder, Watson!" (2) "We'll let people summon other people in cars using their phones." (3) "Okay, so we'll call it 'Bee Movie', because it's about bees."
<Infinisil>
Heh
<simpson>
Seriously, never forget, Bee Movie was written by a bunch of comedians sitting around smoking cannabis. "And they'll have lawsuits against the humans. No, no, lemme explain, this is good."
<Infinisil>
Lol. I won't ever forget the bee movie memes, it's been great
<gchristensen>
cool, I think I got it
<dash>
Infinisil: and remember, if your ideas are any good, you never have to worry about people stealing them
phinxy has quit [(Read error: Connection reset by peer)]
<Infinisil>
dash: s/are/aren't ?
<Infinisil>
gchristensen: What is `lists`?
<gchristensen>
lib/lists.nix
<Infinisil>
Ah
<Infinisil>
Nice!
<simpson>
Infinisil: Famous saying: Don't worry about anybody stealing your ideas. If your ideas are any good, you'll have to cram them down peoples' throats.
<gchristensen>
now comes the hard part, outputting in to the manual somehow
newhoggy has joined #nixos
cornfeedhobo has joined #nixos
<Infinisil>
simpson: Ohhh I know that too well. I've talked a lot about the ideas I'm having with my (kinda) friends, but I quickly stopped that as there was only criticism
<Infinisil>
I'm afraid this might be the norm for people in computer science
<simpson>
Infinisil: If you can't talk with them about stuff in your life, they might not be friends, or you might not know how to talk about stuff.
<Infinisil>
simpson: Might be both heh
hellrazo1 has joined #nixos
<Infinisil>
Did you know, the first guy who wanted to publish the first paper on public key cryptography was harshly dismissed by one of the first reviewers, because the idea of it was so absurd to them that it couldn't work
lambdael has joined #nixos
lambdael has quit [(Client Quit)]
lambdael has joined #nixos
vaibhavsagar has joined #nixos
<Infinisil>
And now we're here, the whole world is using public key cryptography
<simpson>
To be fair, we still don't know whether one-way functions exist.
<simpson>
Humanity tends to prefer beta-quality things that appear to work over proofs which aren't intuitive.
<simpson>
So "this function mixes up your bits a bunch and it isn't invertible" is a claim that humans are predisposed to agree with.
\\\\\\\\\\\ has quit [(Remote host closed the connection)]
<simpson>
After all, look at those bits! They're so mixed! Of *course* it's not invertible~
<gchristensen>
I mean
<gchristensen>
simpson
<gchristensen>
have you seen how mixed they are?
<Infinisil>
Some security protocol was using something like Enc(Hash(Enc(...), ..., ...), years later it was discovered that this extra nesting doesn't do anything and could have been left out. They just used it to be on the safe side, no proof
<Infinisil>
simpson: I feel like quantum computing could somehow be used for that
<sphalerite[m]>
what about more exotic ones like strings with NUL characters inside?
<nixy>
LnL: Not sure I follow entirely, are you saying that attribute names can be anything as long as they are quoted properly?
<LnL>
yes
<clever>
sphalerite[m]: i ran into a problem a few days ago, builtins.readFile cant read a binary
<sphalerite[m]>
aww
<nixy>
Hmm... It looks like symbolic names can contain '.'s too :( This isn't as straightforwad as I hoped
<sphalerite[m]>
clever: is there a way to read binaries in nix at all without import-from-derivation or extending nix?
<clever>
and because of bash, you cant just nix-build -A "foo.bar" or -A foo."bar.baz"
<sphalerite[m]>
surely you can, -A 'foo."bar.baz"'
takle has joined #nixos
<clever>
yeah, and thats not really user friendly
<LnL>
I would expect that to work
<clever>
sphalerite[m]: i dont think there is any way to read a binary string, i was thinking of a new primop to fix the problem there
<sphalerite[m]>
clever: do you have a better solution?
<sphalerite[m]>
(a better solution for command interpretation, not the binary file reading thing)
<clever>
sphalerite[m]: the problem i was having then, is that i needed to force a file in the store to be re-copied to the store, as a content-addressed entry (fixed output)
<clever>
do you have some example code or input and what you want done?
<sphalerite[m]>
no I mean do you have a better solution (alternative shell syntax) for -A 'foo."bar.baz"'
jgertm has joined #nixos
<LnL>
clever: like copyPathToStore?
<clever>
LnL: yeah
<LnL>
builtins.filterSource kind of does what you want then
<jophish>
sphalerite[m]: clever: I'm doing more or less what you suggested, dd ... | ssh root@arria10 -- dd of=/dev/disk/by-label/NIXOS_BOOT (after unmounting /boot), however I seem to be getting kernel panics on boot from the new system
<clever>
LnL: the problem me and taktoa[m] where trying to solve, is to take an unpacked .tar.gz, and then copy every single .c and .h out of it, into seperate storepaths, content addressed
<clever>
LnL: filterSources refuses to accept a storepath as an input
<jophish>
which is odd, as I can remount /boot (before rebooting) and everything appears OK
<sphalerite[m]>
jophish: sorry, I think I'm missing some context
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] volth opened pull request #27398: tinc: allow the daemon to write to files in /etc/tinc/${network}/hosts (master...tinc-fix) https://git.io/vQdsR
NixOS_GitHub has left #nixos []
<sphalerite[m]>
what's the problem you're solving and what did I suggest?
<jophish>
sphalerite[m]: At the moment I'm making lots of sd card images, and dding them to a card, then plugging that into a machine to run from
<sphalerite[m]>
aah right
<sphalerite[m]>
BIOS or EFI?
<jophish>
but I'd like to just copy the changing part (the boot partition) over the network
<jophish>
u-boot
<sphalerite[m]>
ah ok
<jophish>
well, my original question was about copying the whole image
<jophish>
you suggested `curl | dd of=/dev/sda` :D
<sphalerite[m]>
ooooh yes
<jophish>
it's not too often that the rootfs changes, so I'm starting with /boot
jgertm has quit [(Ping timeout: 258 seconds)]
<jophish>
but eventually I'd like to overwrite / properly (I suppose with some tmpfs root)
<sphalerite[m]>
you could also do a sort of blue-green thing like chrome os
<sphalerite[m]>
have two of each of boot and root, then always overwrite the one that's not currently booted
<clever>
but how will the system know which boot to use?
<jophish>
that's a good idea, however this system demands that /boot is partition 1, so I can't have two boot partitions
<TimePath>
do they have two of /boot?
<TimePath>
I thought they had one and just toggled a flag
<TimePath>
clever: with syslinux, it's doable by using `include` on a symlink that points to the active partition
<TimePath>
similar for the others I guess
<clever>
ah, but you would need a 3rd partition for syslinux and that config
<TimePath>
or accept having a mutable /boot
<TimePath>
which should rarely change
jgertm has joined #nixos
<TimePath>
if it does need to, an OS upgrade can always handle it
<sphalerite[m]>
TimePath: well actually it doesn't have a dedicated boot partition at all AFAICT
<clever>
i believe android does it by having a thin bootloader before the real one, that can reflash over usb, and must never change (i think its a rom in the cpu die)
<clever>
TimePath: is this an allwinner cpu your on?
<sphalerite[m]>
instead it has a partition that contains just the kernel and boot options
<sphalerite[m]>
and a partition for the rootfs
<TimePath>
clever: I looked into this for an embedded x86 computer
<sphalerite[m]>
with two copies of each
<TimePath>
nothing flashy
<sphalerite[m]>
jophish: use linux as a second-stage bootloader :p
celldee has quit [(Quit: leaving)]
<jophish>
sphalerite[m]: I've had enough trouble with bootloaders!
<jophish>
but in principle, my script should work. It finds the boot partition in the sdcard image, pipes that across an ssh connection, and the remote side writes it to /dev/disk/by-labe/NIXOS_BOOT
<clever>
basicaly, it will either boot over usb, or it will execute a block at a hard-coded offset on the sd/mmc/nand/spi flash
mudri has quit [(Ping timeout: 240 seconds)]
<clever>
and typically, you put a combined SPL + u-boot at that offset of the block device
<clever>
and then everything else is up to u-boot
jgertm has quit [(Ping timeout: 260 seconds)]
tanonym has quit [(Quit: Going offline, see ya! (www.adiirc.com))]
<ixxie>
is there some option where I can specify a script to be executed on startup?
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<sphalerite>
ixxie: when is "on startup"?
<ixxie>
heh
<ixxie>
good question
<clever>
ixxie: one-shot systemd unit?
<ixxie>
after I already logged into my user account and after gnome is already up and running
<sphalerite[m]>
But what you want is probably to create a OneShot systemd service
<ixxie>
I am, but I was specifically asking for configuration options and that is the only one with mention of xdg
<ixxie>
but I can just go for the normal dotfile approach ^^
Itkovian has quit [(Client Quit)]
<ixxie>
and in anycase people have already advised me here on how to specify dotfiles inside my nix config
Itkovian has joined #nixos
thc202 has joined #nixos
<sphalerite[m]>
ixxie: if you want it for every user, you can also make a little package for it and add it to systemPackages, I think
peel has joined #nixos
<sphalerite[m]>
wait no
tmaekawa has quit [(Read error: Connection reset by peer)]
<sphalerite[m]>
add a .desktop file to /etc/xdg/autostart, you can do that using the environment.etc option
<clever>
and if you add something to systemPackages that has a $out/etc/xdg/autostart, the same thing will work
tmaekawa has joined #nixos
<sphalerite[m]>
oh, ok
<ixxie>
sphalerite[m]: this should really be per-user; what I am aiming for is a way to combine the workspace grid gnome extension with devilspie2 and an autostart script, to launch software in a particular geometric configuration
<ixxie>
but good to know all of that :)
<sphalerite[m]>
clever: that approahc would also work if the package were installed by a user right?
<sphalerite[m]>
ixxie: use a tiling window manager xD
<ixxie>
sphalerite[m]: which would you recommend?
<sphalerite[m]>
well if it's per user I'd go for the package option as mentioned by clever, since that should be installable in a user profile too
<TimePath>
jophish: when working on the embedded x86 project, I ran into similar issues
mpcsh has quit [(Quit: THE NUMERICONS! THEY'RE ATTACKING!)]
mpcsh has joined #nixos
<TimePath>
the flashed image worked, the first update worked, the second update failed
<jophish>
yeah, that sounds like what I'm seeing
<ixxie>
cheers sphalerite[m]
<jophish>
TimePath: although, when I was writing to /dev/disk/by-label/NIXOS_BOOT there was no seek value
<jophish>
so I don't know if seek_bytes would make a difference
<TimePath>
I remember oflag=seek_bytes fixed something, that may have been my build script though
* TimePath
digs up proprietary source he can't really share
hotfuzz has quit [(Read error: Connection reset by peer)]
hotfuzz has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vbgl pushed 1 new commit to master: https://git.io/vQdGb
<NixOS_GitHub>
nixpkgs/master 071693e Vincent Laporte: matio: 1.5.9 -> 1.5.10
NixOS_GitHub has left #nixos []
<TimePath>
jophish: the solution I came up with used /dev/sdaX rather than partx and /dev/sda, and wrote raw partitions straight on top from userspace
ixxie has quit [(Quit: Lost terminal)]
<TimePath>
take from that what you can?
<pie_>
anyone know how to fix the prompt being broken in nix-shell?
<sphalerite>
pie_: are you nesting nix-shells?
<pie_>
sphalerite, no
<pie_>
oddly enough SHLVL=2
<pie_>
whatever that means
<pie_>
thats for my base shell...
<sphalerite[m]>
same for mine
<pie_>
huh ok i guess thats not important
<pie_>
this is what my prompt looks like \[\][nix-shell:/home/wut]$\[\]
Itkovian has joined #nixos
<sphalerite[m]>
does exec bash help?
<pie_>
i think i may have had thisproblem before and it was some env var issue
<sphalerite[m]>
yeah, the issue is very likely to be that it's the minimal bash from stdenv that's running, rather than the full-featured interactive bash
<pie_>
no that doesnt fix it
<pie_>
echo $SHELL
<sphalerite[m]>
find out what the path to your regular bash is and try execing that
<pie_>
/run/current-system/sw/bin/bash
<sphalerite[m]>
or just exec $SHELL in that case
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vcunat pushed 7 new commits to staging: https://git.io/vQdZm
<NixOS_GitHub>
nixpkgs/staging f5a7ce1 Franz Pletz: ldns: also build examples
<NixOS_GitHub>
nixpkgs/staging 3bb9954 Franz Pletz: dns-root-data: init at 2017-06-16
<sphalerite[m]>
compare readlink -f /proc/$$/exe in the broken one and a working one
<srhb>
Are there any good examples of handling (clustered) services requiring the same certificates somewhere in the nixos modules? I gather that what I want to do is initially generate all the certificates on one host and then on others import it using one of the fetch... functions elsewhere
<pie_>
and now it doesnt work (i probably fixed it with something funky)
<pie_>
cant reproduce fix
<sphalerite>
bash is weird
<sphalerite>
env vars are weird
<sphalerite>
computers are weird
* pie_
flails helplessly
<sphalerite[m]>
regardless, `exec /run/current-system/sw/bin/bash` should usually fix it
<pie_>
oook that does fix it
<pie_>
even though exec $SHELL doesnt
<pie_>
which is weird because they are the same
<pie_>
well ok not after i just installed bash globally
<pie_>
which you would expect to fix it???
<sphalerite[m]>
no
<sphalerite[m]>
that's what breaks it
<pie_>
...
<pie_>
wtf?
<pie_>
(i did that after we started talking, for the record)
<sphalerite[m]>
bash is the one that doesn't work. bashInteractive is installed on the system by default.
<sphalerite[m]>
bashInteractive is the one that you want.
<pie_>
oh.
<pie_>
:I
<pie_>
thats unexpected.
<pie_>
and weird. why is it like that?
<pie_>
yay it works \o/
<sphalerite[m]>
because for the purpose of building packages, bash without interactive is usually what you want because it has a much smaller closure
<pie_>
(FINALLY using a shell doesnt feel like crap!)
<jophish>
TimePath: thanks, I'll give that another try
<pie_>
sphalerite[m], thanks for the help!
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<sphalerite[m]>
pie_: it's weird though, it should use bashInteractive for nix-shell by default
<sphalerite[m]>
if it's installed systemwide at least
<sphalerite[m]>
I think
<pie_>
unrelated: is the [m] some kind of special irc suffix here?
<srhb>
I can generate certificates using eg runCommand to build them, and presumably I can deploy these with nix-copy-closure, but how do I refer to them in configurations on the hosts they were deployed on?
<srhb>
Dealing with impurity is hard :-)
orivej has quit [(Ping timeout: 240 seconds)]
takle has joined #nixos
<TimePath>
impurely refer to /var?
<srhb>
So generate them with runCommand, stick them in /var and just transfer them around manually?
<srhb>
I guess I could do that...
gpyh has joined #nixos
<gpyh>
hi
<srhb>
gpyh: Hi.
<gpyh>
I'm updating my system right now and I have a dependency on snappy that raises an error 500 from the cache
<gpyh>
weird stuff
<gpyh>
anybody having the same problem?
<clever>
srhb: if sandboxing is on (it should always be on), there is no way for runCommand to access /var/
<srhb>
clever: Right.. So this basically has to be dealt with outside of Nix
<clever>
srhb: so the result of runCommand will wind up in the store (with world-readable keys), and you then need to use nix to insert that storepath somewhere
<gpyh>
It seems to just be a problem with snappy because once I removed chromium (which depends on it) it updated fine
<srhb>
gpyh: What's the hash?
<aristid>
gpyh: i also had error 500 issues with the cache earlier today
<srhb>
gpyh: Or which nixpkgs
<gpyh>
last unstable
<aristid>
gpyh: i would assume/hope it will fix itself in a few hours
<srhb>
gpyh: Works for me. Guess it's transient?
<gpyh>
ok thanks ;)
<gpyh>
srhb: maybe, I'll try again later this day
<gpyh>
errors 500 are concerning though
<srhb>
clever: Yeah that doesn't sound nice..
[0x4A6F] has quit [(Remote host closed the connection)]
<aristid>
i think cache.nixos.org is cloudfront
[0x4A6F] has joined #nixos
iyzsong has joined #nixos
<LnL>
yes s3 + cloudfront so it "should" just work
simukis has quit [(Remote host closed the connection)]
<LnL>
clever: srhb: it would be interesting if it was possible to create a store that's only accessible by the current user
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vcunat pushed 4 new commits to staging: https://git.io/vQdng
<NixOS_GitHub>
nixpkgs/staging 5a03a05 Vladimír Čunát: gtk3: maintenance 3.22.15 -> 3.22.16
<NixOS_GitHub>
nixpkgs/staging 7d80f94 Vladimír Čunát: dbus: maintenance 1.10.18 -> 1.10.20
<NixOS_GitHub>
nixpkgs/staging e2b7b09 Vladimír Čunát: Merge branch 'master' into staging
NixOS_GitHub has left #nixos []
<srhb>
LnL: That would do some if it, yeah.
simukis has joined #nixos
ibrahims has joined #nixos
<srhb>
LnL: I looked some more into the kubernetes issues. I think your upversioning might be closer to an actual release, since it's less different from the current one. Are you using it currently, or did you leave the effort in favor of matejcs PR?
<LnL>
yes I picked the versions based on what's used by the release
<srhb>
I think 25426 might have stagnated though..
<LnL>
but I basically don't know anything about k8s
<LnL>
was just playing with it and noticed it was a bit outdated
* srhb
nods
<srhb>
offlinehacker[m]: Do you have any idea of current progress on k8s upversioning? PR #25426 seems to have stagnated for a bit, is anything happening behind the scenes, or would it be preferable to work on upversioning the current source build?
[0x4A6F] has quit [(Ping timeout: 246 seconds)]
<LnL>
so 25426 is a bit complicated for me to review or use as a starting point
<srhb>
LnL: Same sense here, also very limited experience.
<srhb>
(Currently evaluating going to Kubernetes instead of Mesos)
[0x4A6F] has joined #nixos
newhoggy has joined #nixos
orivej has joined #nixos
newhoggy has quit [(Ping timeout: 268 seconds)]
takle has quit [(Remote host closed the connection)]
<gpyh>
cache issue seems to be fixed
<gpyh>
great :D
necronian has joined #nixos
<sphalerite[m]>
LnL: it is possible to create a store that's only accessible by the current user
<sphalerite[m]>
e.g. using nix-user-chroot. It's a bit of a pain to use on nixos though because you need to somehow ensure that the closure of current-system and ~/.nix-profile are available in it
<sphalerite[m]>
or get lots of command not founds :p
<simukis>
installer ignores --max-jobs and --cores and downloads stuff sequentially
andymandias has quit [(Ping timeout: 268 seconds)]
<srhb>
simukis: nixos-install ?
<simukis>
yes
<simukis>
now I’m waiting for it to download my largeish config package-by-package
<srhb>
Sounds issue-worthy
instantepiphany has joined #nixos
Drakonis[m] has quit [(Changing host)]
Drakonis[m] has joined #nixos
Drakonis[m] has quit [(Changing host)]
Drakonis[m] has joined #nixos
andymandias has joined #nixos
<instantepiphany>
I just made my system unbootable :( I tried using the AMD proprietary drivers, by adding the right line to my configuration. After rebooting, X failed to start. I didn't have time to troubleshoot the error, so I just chose my previous configuration (8). Booted fine as you would expect. Just now I wanted to make that boot the default, so I ran `sudo
<instantepiphany>
nixos-rebuild boot --rollback`. The output showed it rolled back successfully, but now I can't boot, I get a bunch of EFI errors and when it shows the nixos grub screen, I can only see "Nixos" and not any of my profiles. What gives?
<instantepiphany>
I am in the live cd for nixos now on the same machine, what logs should I look at to track this down?
<instantepiphany>
I am surprised that running the rollback flag rendered my system unbootable, I haven't made any changes ever to my boot setup on this install, only what GPU drivers are being used.
<LnL>
sphalerite: oh really?
<LnL>
doesn't nix always make paths world writable
<LnL>
*readable
<gchristensen>
domenkozar: can you test 1f.11.13?
<srhb>
instantepiphany: That sounds bad. Can we see the EFI errors?
<vaibhavsagar>
hey, I'm trying to build a package with stack --nix and it's complaining about a missing ncurses
<instantepiphany>
srhb: Sure, whats the best way for me to capture them? They only show for a second, and then the next boot device is tried until the boot devices are exhausted.
<instantepiphany>
I could just take a photo?
<vaibhavsagar>
I've added 'ncurses', 'ncurses.dev', and 'ncurses.out' to my stack.yaml
<srhb>
instantepiphany: That sounds like the "easy" solution, yeah.
<instantepiphany>
Ok, back soon.
<vaibhavsagar>
it looks like other people have had this issue before?
<sphalerite[m]>
LnL: sure, but if the store is in a path that's inaccessible to other users (e.g. in your home and your home is go-rx) they can't access it
<LnL>
slyfox: cannot read file system information for '%u': No such file or directory
<slyfox>
thanks!
<LnL>
slyfox: stat -c works in a drv
<slyfox>
*nod*
HoloIRCUser has quit [(Read error: Connection reset by peer)]
<LnL>
because we use the gnu coreutils for builds
k0001 has joined #nixos
<LnL>
sphalerite[m]: ah like that, I don't think things like ssh are happy with that
ng0 has joined #nixos
pie_ has quit [(Ping timeout: 240 seconds)]
<sphalerite[m]>
LnL: then have it in ~/private. Whatever. You get the idea
<LnL>
yeah, but the other ideas are harder to implement
<LnL>
and support on every platform
<gchristensen>
true
<instantepiphany>
srhb: I think those might be from my motherboards uefi booter. Does that mean nixos-rebuild somehow corrupted the boot partition?
<LnL>
having a separate secret store with derivations that can't be substituted is a lot simpler
pie_ has joined #nixos
pie__ has joined #nixos
pie__ has quit [(Remote host closed the connection)]
<gchristensen>
that does fix some problems, since they're only readable by your user
<LnL>
but yes authorisation/auditing would be nice
eacameron has joined #nixos
<srhb>
instantepiphany: It looks like perhaps the kernel is missing?
<simukis>
install finally done
<srhb>
instantepiphany: Do you see it in /boot/EFI/nixos ?
<LnL>
gchristensen: btw, I really like your function docs idea
<gchristensen>
oh really? :o
<gchristensen>
"It does make the generated documentation nice to read, but less so the source itself." I agree with the end of this a lot
<LnL>
one thing that doesn't really work but that would require some support by nix is actually annotating the functions somehow
<instantepiphany>
srhb: the efi file that contains the kernel version? Yes that is present in the directory you mentioned. Along with 2 initrd efi files.
<gchristensen>
right
<LnL>
eg if you have f = ({ meta.type = "id :: a -> a"; __functor = self: id; })
<gchristensen>
LnL: nix is so dynamic though, there is no function table to go look at
<LnL>
then nix-repl's :t you lookup f.meta.type and print that
<srhb>
instantepiphany: OK, so at least it's trying to find the right one. Is there anything suspicious about it? Filesize look OK?
<srhb>
instantepiphany: What its modified time?
<MoreTea>
LnL, I added a :h command to the nix-repl
<MoreTea>
There are two approaches that I worked on
<LnL>
oh!
<MoreTea>
1) one that adds doc-comments to functions in nix, and a new built-in like builtins.getFunctionArgs
<instantepiphany>
srhb: bingo! All 3 efi are 0 bytes. All have a modification time that would match when I ran the roll back.
<MoreTea>
2) one that abuses __functor to fake this behaviour
<instantepiphany>
srhb: if I had 8 profiles do they live in this folder as efi files or is the correct profile just passed in as an argument for the os to handle?
<gchristensen>
a common problem to all three of these is finding the things which have docs, and including them in the doc output
<gchristensen>
(doc output == the manual)
<MoreTea>
this would allow us to write documentation in a nested way; e.g. document lib.strings and lib.strings.someFunction separately.
<MoreTea>
gchristensen, we can crawl through all nested attrsets in nixpkgs, and find documentation.
<srhb>
instantepiphany: I think they just represent unique kernel/initrd combinations in the currently live generations, so it's <= 8 pairs for you
<srhb>
instantepiphany: (and the rest is arguments from grub)
<gchristensen>
MoreTea: do you have ideas on how we'd know it came from lib.strings and not foo.lib.strings?
<MoreTea>
yes, I posted you the link to the generated manual, right?
<srhb>
instantepiphany: And it sounds like the builder was somehow interrupted while writing them, yikes :/
<instantepiphany>
Ok well I never changed any kernel parameters so that makes sense, apart from the 0 size. Is there a way for me to recover from the live cd?
<srhb>
instantepiphany: You should be able to just rebuild boot though
<MoreTea>
it is based on the __functor trick, but can equally well be adjusted for the builtins approach.
<gchristensen>
MoreTea: nice ...!
<gchristensen>
how do you do that?
<instantepiphany>
srhb: just so you know, when I ran the builder I had no error messages at all, it just had a normal looking message like "completed.". How do I rebuild from live though? If you don't know off the top of your head I'll just go looking, no problem.
erictapen has joined #nixos
<srhb>
instantepiphany: I think you can just chroot and nixos-rebuild boot
<MoreTea>
have a function that iterates with attrNames over an attr set, then recursively calling the same function, with one extra argument: the prefix of the path that it took to get there.
<clever>
gchristensen: does your new darwin thing put nix into PATH early enough for "ssh darwin nix-store --version" to work?
<MoreTea>
oh, nice.
<MoreTea>
I don't know any lisp though :x
<gchristensen>
MoreTea: me either
<gchristensen>
oh beautiful, have fun :D
<nixy>
I have been in a book club for lisp for three months and I still don't know lisp :y
<gchristensen>
MoreTea: but I was discussing the challenges of docing nix with a friend who does know lisp, see the top of myPR :)
<instantepiphany>
I have my system set to default channel (stable) but my user set to use unstable. So nixos-install --chroot should work?
<instantepiphany>
I'll give it a go.
<srhb>
instantepiphany: That should be perfectly fine
<MoreTea>
nixy, haha, it's one of those minimal systems that appears to be simple
<srhb>
instantepiphany: Worst case is you install a slightly different version of nixpkgs, and I don't think that will happen with your channel setup
<srhb>
instantepiphany: (in fact, you could have followed the installer verbatim safely, but we're shaving off a few steps with chroot)
<gchristensen>
MoreTea: tbh I love the functor PR
<srhb>
Well, aside from the formatting ;-)
<MoreTea>
I like that someone benchmarked the functors, and found out that it's only 3% slower
<gchristensen>
MoreTea: what was the impact in how long it takes to make the docs?
<instantepiphany>
Ok that worked instantly, cool. Now just run nixos-rebuild?
<MoreTea>
I edited the existing documentation to fit this pattern
<srhb>
instantepiphany: If you have your boot mounted in your new root, yes
<gchristensen>
I mean
<gchristensen>
`nix-build` the docs
<instantepiphany>
And that should get me back to where I am, and then a roll back like I did in the first place is still what I should do. And on the off chance this happens again now I know how to fix.
<LnL>
MoreTea: well question is if that's just for the docs or "every" call of a documented function
<MoreTea>
oh, given that it only documents lib.*, it's a few seconds
<srhb>
instantepiphany: There will be no need to roll back unless you still have the broken radeon things in your configuration.nix
<srhb>
instantepiphany: In which case I'm not entirely sure where rollback will bring you now
<MoreTea>
LnL, every call to a lib function.
<gchristensen>
MoreTea: oh that is a lot :( maybe improvable though.
<srhb>
instantepiphany: Easier to just make sure your config doesn't have the broken things and rebuild a new generation with that
<MoreTea>
gchristensen, if this were a build-in function, yes.
<LnL>
yeah that's why I was thinking that maybe adding function metadata to nix might be better
<gchristensen>
yeah but then we're moored to whatnix can do
<srhb>
instantepiphany: As a side note, I wish I knew what logs to tell you to post as an issue for this, because it's a really bad thing to happen, but I sadly don't :/
b has joined #nixos
<MoreTea>
gchristensen, I'm willing to play a long game here.
<MoreTea>
I thought about doc-comments, but maybe a wrapper function that does what my __functor PR does is more acceptable.
<MoreTea>
Although I think that these doc-comments can also open up the doors to opt-in type checking
<instantepiphany>
srhb: Good point. I got 2 errors when running rebuild. Parent directory /run/user/0 does not exist at <nixstorepathfornix1.11>
<gchristensen>
whoa now buddy :P
<instantepiphany>
Yeah I wish I knew the right logs too.
<srhb>
instantepiphany: err umm.
<gchristensen>
yeah, I agree, MoreTea.
<gchristensen>
instantepiphany: don't run nixos-rebuild in a nix-shell
<LnL>
what was the issue with that again?
<gchristensen>
I dunno
<LnL>
there's some kind of bug, but I don't remember what
<srhb>
gchristensen: Why would nixos-install --chroot be a nix shell?
<instantepiphany>
Does running nixos-install drop you in a nix-shell?
<srhb>
It sounds more like it didn't mount in /run
<MoreTea>
I would really like to have "jump-to-definition" in Nix files, or "show me the documentation for the function under my cursor"
<gchristensen>
oh, sorry, I assumed you had just run nix-shell on your own
<instantepiphany>
No, but thanks!
<srhb>
instantepiphany: If you exit the chroot and simply nixos-install --chroot nixos-rebuild boot, does that work?
<srhb>
I don't think it should either, really
<srhb>
It sounds like it's missing some mounts that I assumed it would make for you
<instantepiphany>
I'll try that
<gchristensen>
:) MoreTea: yeah, my PR makes progress on that with unsafeGetAttrPos but I'm having a hard time finding where that would go in your PR
<srhb>
instantepiphany: If that doesn't work I think it may be easier to just run nixos-install regularly, but I'm not sure whether I should recommend that you back up things first.
<srhb>
You could do all the mounts by hand, but ugh.
eacameron has quit [(Remote host closed the connection)]
<instantepiphany>
I ran your last suggestion and got the same errors which is strange as the directory clearly does exist.
<srhb>
instantepiphany: Inside the chroot as well?
<ToxicFrog>
does nix-shell -p not respect overrides?
<srhb>
instantepiphany: as in, /mnt/run/user0 ?
<srhb>
ToxicFrog: Yes.
<ToxicFrog>
...yes it does respect them, or yes it ignores them?
<srhb>
ToxicFrog: Yes it does. :-P
<gchristensen>
ToxicFrog: it'll respect overrides in your user's nixpkgs config
<gchristensen>
but not in your global nixos config
<MoreTea>
gchristensen, I like your approach of splitting it into a 'docs' attrset.
<instantepiphany>
Oh those errors are from inside the chroot. No, /mnt/run is empty.
<srhb>
instantepiphany: Yeah, I expect it should be bind mounted in
<gchristensen>
MoreTea: oh you do?
<gchristensen>
MoreTea: one huge idea you just proposed was optional type rules, which are not possible with my implementation
<MoreTea>
correct
<gchristensen>
which I thought was just incredible.
<MoreTea>
the advantage of your approach is no loss in performance.
<gchristensen>
what if I told you we could have our cake and eat it too
<MoreTea>
It's more a random thought, than something grand plan
<MoreTea>
hmm?
<ToxicFrog>
gchristensen: thank you. That explains the problem.
<instantepiphany>
srhb: bind mount /run to /mnt/run?
<gchristensen>
what if we used your mkDoc / functor approach, but by default when "exporting" the attrset we (by default) stripped the docs out and returned the inner functions
<gchristensen>
s/mkDoc/docFn/
<srhb>
instantepiphany: That's what I was thinking. Though to be honest I'm starting to feel this is getting more sketchy than just running nixos-install
<gchristensen>
ToxicFrog: you're welcome, glad to help :)
<srhb>
instantepiphany: I think the problem might be that it cannot get nix-daemon running.
<MoreTea>
gchristensen, I'm not sure of that thunk will be evaluated just once
<MoreTea>
*if
<gchristensen>
MoreTea: that way call-time people are accessing the inner fn directly, but perhaps an import-time parameter lets you keep them, or even the docs portion gets duplicated over to { docs.___ = }
<gchristensen>
hmm yes
<ToxicFrog>
Hmm. So now I have the override in place, I'm installing peruse as part of systemPackages, and /nix/store/...-peruse-1.2+20170127/ is getting created
<ToxicFrog>
But looking inside it, bin/peruse is just a bash script that forwards to peruse-1.2/bin/peruse
<ToxicFrog>
o.O
<MoreTea>
we could test this in a simple PoC with a builtins.trace, to see if a "to-doc-or-not-to-doc"-flag is evaluated once, or every time it's used.
<MoreTea>
anyhow, my time is up for now.
<ToxicFrog>
All I did was overrideDerivation with a new src attribute
<instantepiphany>
srhb: Well nixos-install should not remove files from the file system should it? I can use my existing config, without the radeon lines. All I should have to do after the install is run my nix-env stuff for my user profile again, right?
<gchristensen>
continuing down this thought, despite your very good question, an interesting thing about this is then the docFn could wrap up the function with type checking, if an import-time parameter says to
<MoreTea>
yesss
<srhb>
instantepiphany: I'm not even sure that will be necessary, but yes.
<MoreTea>
ttyl!
MoreTea has quit [(Quit: Leaving)]
<gchristensen>
MoreTea: see you!
eacameron has joined #nixos
<instantepiphany>
I just want to get my system working as quickly as possible, I'm not really sure what's involved in replicating the right parts of nixos-install
<ToxicFrog>
As far as I can tell it doesn't attempt to build or even download the new source, it just emits this wrapper around the old source o.O
<ToxicFrog>
Doing this worked fine for other overrides...
<instantepiphany>
shrb: What do you think would be faster?
<srhb>
instantepiphany: After your mounts are correct and your /mnt/etc/nixos/configuration.nix is correct, just nixos-install
<gchristensen>
ToxicFrog: can you paste your config?
newhoggy has quit [(Remote host closed the connection)]
<instantepiphany>
srhb: running nixos-install I get errors about the boot partition being read only.... I didn't mount with that flag.
<ToxicFrog>
All the overrides there work as expected except peruse
<gchristensen>
ToxicFrog: where does this live?
<srhb>
instantepiphany: remount rw
<ToxicFrog>
/etc/nixos/common/overrides.nix and it's imported with imports = [ ./common/overrides.nix ] in configuration.nix
eacameron has quit [(Ping timeout: 240 seconds)]
<gchristensen>
ok, and you're not testing with nix-shell?
<instantepiphany>
I just remounted and it had rw flag, that means it must have errored during mounting before.
<ToxicFrog>
Not right now, no
<ToxicFrog>
I do `nixos-rebuild switch` and it emits a directory in /nix/store for the version in the override, but it just contains a wrapper script
<srhb>
instantepiphany: How did you mount it originally? mount /dev/sdXY /mnt/boot ?
<gchristensen>
ToxicFrog: might want to use overrideAttrs instead of overrideDerivation
<instantepiphany>
srhb: it never ends haha! I now get oserror for /boot/loader/entries/nixos-generation-3.conf.tmp being read only.
<instantepiphany>
Yeah that's how I mounted it
<srhb>
instantepiphany: You might want to just nuke your /mnt/boot/* from orbit
<gchristensen>
I'm not sure why it'd make a difference here, but I dn't know
<srhb>
instantepiphany: Since it's useless anyway.
<instantepiphany>
srhb: that's what I was thinking.
<ToxicFrog>
gchristensen: looking at /nix/store, it looks like even without the override it emits *three* derivations
<ToxicFrog>
One is a wrapper script
<gchristensen>
wtf?
<ToxicFrog>
Thta calls a symlink in the second one
<ToxicFrog>
Which points to the third one which contains the actual binary
<gchristensen>
ohhh
<instantepiphany>
srhb: the plot thickens! I checked mount, and the boot partition is ro. When I mounted it, it was rw. Could nixos-install somehow have set it to ro?
<ToxicFrog>
like, running `peruse` calls /nix/store/wv3lnmcg7m2013nrk41ribln2li13b20-peruse-1.2/bin/peruse, which calls /nix/store/ldianvkcflnmnzhnwirkh3pdalzd9v43-peruse-1.2/bin/peruse, which is a symlink to /nix/store/kwkxf7r3kc232fg0y48s1azw2h2w9vz1-peruse-1.2/bin/peruse, which is the actual program I'm trying to update
<srhb>
instantepiphany: I honestly don't know, but since I know it can install given the right circumstances, it must be rw at some point ;-)
newhoggy has quit [(Ping timeout: 248 seconds)]
<srhb>
instantepiphany: I still think you should nuke it and try again rw
<ToxicFrog>
gchristensen: where does that show up? A nixpkgs repo search doesn't find it
<gchristensen>
are you using github search?
<ToxicFrog>
Yeah
<gchristensen>
all-packages.nix is excluded from their index b/c it is too long
<srhb>
instantepiphany: Though, if the partition itself is in bad shape, I suppose it's plausible that it gets remounted ro by the kernel itself
<ToxicFrog>
thanks for nothing, github search
<gchristensen>
right? it lies to you exactly when you need search
<srhb>
instantepiphany: That could be interesting to look into if you have the time.
<srhb>
instantepiphany: Since it might also explain the earlier failure.
<ToxicFrog>
It also doesn't index non-alphanumeric characters, which makes it particularly bad for searching code
<ToxicFrog>
Which is baffling on a site that exists to host code
<instantepiphany>
srhb: yeah but now I am expecting hardware issues. It is set to remount ro on error. When I tried rm it wouldn't touch the generation.conf files, it complained read only file system.
<instantepiphany>
Maybe I set the partition up wrong.
<gchristensen>
ToxicFrog: it is a shockingly bad search tool
<gchristensen>
ToxicFrog: on the bright side, search.nix.gsc.io has editor integration!
<instantepiphany>
srhb: installation finished.
<srhb>
instantepiphany: Hooray!
<gchristensen>
instantepiphany, srhb what went wrong in this installation?
<srhb>
gchristensen: After a rollback, all files in /boot/EFI were 0 bytes
<gchristensen>
oops.
<srhb>
gchristensen: Ooops indeed. :-P
<ToxicFrog>
So how do I override a Qt5 program? I was kind of hoping not to have to host an entire local fork of nixpkgs.
<gchristensen>
ToxicFrog: digging
<ToxicFrog>
The nixpkgs manual is, as always, unhelpful. :/
<gchristensen>
srhb, instantepiphany: were you at version 1 and then tried to roll back?
<srhb>
I don't think it was version 1 instantepiphany? But it was a boot into an earlier generation than latest followed by a rollback
<instantepiphany>
srhb: gchristensen: version 1 of my generations? No, went from 8 => 7
<gchristensen>
wow
<srhb>
Yeah, bad.
<instantepiphany>
But surely we don't let users roll back past 1!
<gchristensen>
instantepiphany: well I was hoping we make that silly mistake, vs. something harder to fix :P
<srhb>
There might be some hardware issue at fault
<srhb>
but it would be really nice to know which logs should be salvaged
<srhb>
Because it's definitely slightly terrifying :-P
takle has joined #nixos
<srhb>
Where do rebuild logs live?
<gchristensen>
I don't know that there are any
<instantepiphany>
Should I roll back again to see if it was a God ray?
<gchristensen>
ToxicFrog: ok so I don't know, but I think you don't need all of nixpkgs, but just the one peruse default.nix
<gchristensen>
ToxicFrog: :|
<jophish>
Dezgeg[m]: Do you remember what happened to the "armv7" builds on aarch64 patch you wrote. I'm still using my uname patch and I'm wondering if that's necessary on a recent nixpkgs
<srhb>
instantepiphany: Did it boot up nicely?
<ToxicFrog>
gchristensen: well, that's not great, but it's tractable. Thanks.
smc9115 has joined #nixos
<instantepiphany>
Yeah booted fine
<jophish>
sorry for asking twice :) I might have missed your reply last time
<srhb>
instantepiphany: Might give it a go then.
<instantepiphany>
Apart from x, I booted gen 8
<srhb>
instantepiphany: Sorry it took a while to get you up and running
<instantepiphany>
srhb: no problem I really appreciate the help! To confirm, correct roll back is nixos-rebuild boot --rollback?
<gchristensen>
ToxicFrog: so many things to document, so little time, and so much RSI
<srhb>
instantepiphany: Correct
seif[m] has quit [(Ping timeout: 240 seconds)]
baconicsynergy[m has quit [(Ping timeout: 240 seconds)]
lfont[m] has quit [(Ping timeout: 240 seconds)]
admin[m] has quit [(Ping timeout: 240 seconds)]
Khorne[m] has quit [(Ping timeout: 240 seconds)]
myklam[m] has quit [(Ping timeout: 240 seconds)]
BurNiinTRee[m] has quit [(Ping timeout: 240 seconds)]
jascot[m] has quit [(Ping timeout: 240 seconds)]
pstn has quit [(Ping timeout: 240 seconds)]
sirius[m] has quit [(Ping timeout: 240 seconds)]
offlinehacker[m] has quit [(Ping timeout: 240 seconds)]
Ralith has quit [(Ping timeout: 240 seconds)]
chominist[m] has quit [(Ping timeout: 246 seconds)]
mtncoder[m] has quit [(Ping timeout: 240 seconds)]
rycee[m] has quit [(Ping timeout: 258 seconds)]
alain[m] has quit [(Ping timeout: 258 seconds)]
wmertens[m] has quit [(Ping timeout: 258 seconds)]
frio[m] has quit [(Ping timeout: 258 seconds)]
<instantepiphany>
If I run that command in generation x, does it set the default boot to x - 1, or just remove all generations with n >= x?
<gchristensen>
jophish: Dezge\g was so unhappy with you asking twice they took down all of matrix
<clever>
gchristensen: does your new darwin thing put nix into PATH early enough for "ssh darwin nix-store --version" to work?
<srhb>
instantepiphany: I actually don't know. I can never remember three digit numbers for as long as it takes to boot. :-P
<gchristensen>
clever: hmm
<instantepiphany>
srhb: rofl
<gchristensen>
clever: I could test that for you
<srhb>
instantepiphany: I've been meaning to figure out though!
<srhb>
And also perhaps improve the docs on that ;-)
newhoggy has joined #nixos
magnetophon1 has quit [(Ping timeout: 260 seconds)]
phinxy has joined #nixos
<srhb>
uh
<srhb>
Well, sometimes it's nice when people have possible device failures, because it just made me realize I've lost one disk in my mirror raid without noticing
<gchristensen>
if it is a trivial execable expression
<clever>
handy
<gchristensen>
:)
<ToxicFrog>
Hmm.
<gchristensen>
more like confusing when you see this crazypants pstree and can't figure out wtf is going on and decide to read bash's source at 3am b/c you can't figure how your bashrc is loaded...
<ToxicFrog>
So now I have it building, but for some reason the wrapper emitted doesn't set QML_IMPORT_PATH and QML2_IMPORT_PATH like the one in nixpkgs does
<ToxicFrog>
Which means it can't find its components on startup
<gchristensen>
ToxicFrog: what does your override look like now?
<clever>
gchristensen: after a glance at the bash manpage, i see no mention of /etc/bashrc!
spinus has joined #nixos
<ToxicFrog>
I ditched the override and just load a separate file with pkgs.libsForQt5.callPackage ./common/pkgs/peruse.nix {}
<clever>
so, it only ever runs ~/.bashrc, and /etc/bashrc is just an addition by the distro, (via /etc/profile) ?
<clever>
gchristensen: i also see a /etc/bash.bashrc in the source, hmmm
<olejorgenb>
clever: it's not here (echo $PKG_CONFIG_PATH --> /home/ole/.nix-profile/lib/pkgconfig:/nix/var/nix/profiles/default/lib/pkgconfig:/run/current-system/sw/lib/pkgconfig)
newhoggy has joined #nixos
<olejorgenb>
(Nothing in my personal shell config touches the variable)
<clever>
olejorgenb: ah, wasnt expecting that
<clever>
gchristensen: and on the mac, it ignores /etc/bash.bashrc
<gchristensen>
clever: mac's bash shell is old
hyphon81 has joined #nixos
<clever>
gchristensen: do you happen to remember how to trace syscalls and such on darwin?
<gchristensen>
dtrace but I don't know how to use it
<gchristensen>
and it doesn't work if system integrity protection is enabled
<clever>
i'll google it some more
ebzzry has quit [(Ping timeout: 248 seconds)]
<ToxicFrog>
If I `import <nixpkgs>` in nix-repl, does that get me nixpkgs-with-overrides or just the plain version? If the latter, is there a way to explore with nix-repl what the result of the overrides being applied is?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] rnhmjoj opened pull request #27403: nginx: make listen addresses configurable (master...nginx) https://git.io/vQd4p
NixOS_GitHub has left #nixos []
<clever>
ToxicFrog: if you pass it {}, then it will load ~/.nixpkgs/config.nix automatically
<ToxicFrog>
But not /etc/nixos/configuration.nix?
<clever>
correct, nixpkgs will never load configuration.nix
<gchristensen>
ugh I don't like that there are some things written in markdown and some written in docbook, it leaves me feeling half jealous and half unhappy that I can't copy-paste what they did :P
<clever>
only nixos loads configuration.nix
newhoggy has quit [(Ping timeout: 260 seconds)]
<clever>
ToxicFrog: one handy trick, put all of your overrides into /etc/nixos/config.nix, then in configuration do nixpkgs.config = import ./config.nix;
<clever>
now the config is broken out into a smaller file, that nixpkgs can load on its own
<clever>
gchristensen: looks a lot more powerful then strace
<ToxicFrog>
I had the idea of overriding peruse.unwrapped, as in overrides = pkgs: { peruse = pkgs.peruse.overrideAttrs (old: { unwrapped = old.unwrapped.overrideAttrs (...) }) }
<ToxicFrog>
But that just gives me "value is a list while a set was expected" :(
<clever>
i dont see an unwrapped attribute on my end
<ToxicFrog>
Ok, using the version in unstable as the target of the override seems to be working.
<ToxicFrog>
I have this horrible feeling that I'm going to end up with two versions of KDE installed or something though.
<socksy>
can i ask for some heretical advice?
<socksy>
i often work with build systems that would require a lot of effort to port to nixos, but nixos is my main OS on my dev laptop
<ToxicFrog>
...except it doesn't actually build. argh
<gchristensen>
go ahead socksy
<socksy>
yesterday i found myself writing a Vagrantfile in order to compile some code someone wrote, because the effort for getting the packages with nixos was proving too much and i just need to run and compile once and never work on this POS code again
<clever>
gchristensen: aha, opensnoop looks good
mrcheeks has joined #nixos
<socksy>
isn't there a better way? some kind of temporary, FHS, mutable environment?
<socksy>
one that i can throw away afterwards?
<gchristensen>
socksy: well nix can make a fhs environment
newhoggy has quit [(Ping timeout: 260 seconds)]
<socksy>
i saw that earlier, and i am having trouble with it
<gchristensen>
some people have good luck with the steam tools, which seems to provide a whole distro worth of libs :P maybe check that out?
<gchristensen>
steam-run IIRC
<socksy>
i tried buildFHSUserEnv in a bare nixfile and using nix-shell with it
<socksy>
but i think i must be doing something wrong because it doesn't seem to give me a real FHS env, and besides, it's sim linking to /nix/store which is mounted read only
<clever>
i dont think nix-shell works on that
<gchristensen>
no, I think it provides an executable that you clal
<gchristensen>
nix-build it
<clever>
i think you have to nix-build it, and then it yeah that
<socksy>
ok
<socksy>
how?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vcunat pushed 1 new commit to staging: https://git.io/vQdB6
<NixOS_GitHub>
nixpkgs/staging b3729eb Vladimír Čunát: libuv: 1.12.0 -> 1.13.1
NixOS_GitHub has left #nixos []
<clever>
gchristensen: opensnoop confirms, bash only ever tries to use ~/.bashrc, so it requires imperative setup to work
<gchristensen>
aye
<gchristensen>
IMO we fix this with documentation
<clever>
yeah, no real way to fix this system wide, other then spamage in every home
<clever>
that explains why tab complete has been wonky
newhoggy has joined #nixos
<ToxicFrog>
Ok, it looks like kirigami doesn't build in unstable :/
ibrahims has joined #nixos
<gchristensen>
I'm writing some docs for bundlerEnv since I've used it a lot recently, and there are some cool things in there!
<gchristensen>
drvName =
<gchristensen>
if name != null then name
<gchristensen>
else if pname != null then "${toString pname}-${mainGem.version}"
andreabedini has joined #nixos
newhoggy has quit [(Ping timeout: 268 seconds)]
* ToxicFrog
bangs head against keyboard
<andreabedini>
hello! I am trying to install nix on my mac and I am having some problems. The installes instists in doing the multi-user install (it says 'switching to multi-user installation)
<andreabedini>
how do I tell the installer I want to do the single user install?
<andreabedini>
I got it installed on my other mac a while ago and I remember I was fine with a single user installation
<gchristensen>
andreabedini: may I ask why you want the single-user install? it can be a bit problematic, and is easier to support
<gchristensen>
I think I know why you're having a problem with the multi-user installer, if that is why you want to go back
<gchristensen>
and is harder to support*
<andreabedini>
gchristensen: right, good question, only because I remeber I was easier, perhaps setting up the permission was easier (just mkdir & chmod /nix)
<gchristensen>
the multi-user installer will handle setting up all the perms for you
<andreabedini>
gchristensen: I didn't explain myself too clearly, the problem with the multiuser install is that I (think I) want a single user install
<andreabedini>
I can go with multiuser if it's the best thing to do
<gchristensen>
why do you think you want to? (You may be right! but I'd like to know why, just to understand)
<andreabedini>
gchristensen: less invasive permissions to set up, no extra users or build daemons
proteusguy has quit [(Ping timeout: 246 seconds)]
<andreabedini>
just sounds easier to mkdir /nix and chown $USER /nix
<gchristensen>
ah, fair
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bjornfor pushed 1 new commit to master: https://git.io/vQdRo
<NixOS_GitHub>
nixpkgs/master 9906e44 Stefan Lau: dunst: 1.1.0 -> 1.2.0
<gchristensen>
so a problem with the single user install is it is very easy for (for example) pip or bundler to accidentally write to your nix store, breaking the promises of nix
<clever>
andreabedini: the old script would use sudo to chown /nix for you, and then do the rest without root
newhoggy has joined #nixos
<gchristensen>
is why we are trying to move users to the multi-user version
<andreabedini>
gchristensen: aah, I see ...
<andreabedini>
clever: that's what I remember from installing on the other mac!
* andreabedini
rejoices as the multi-user installation finished without any problems in few seconds
<gchristensen>
andreabedini: my hope was the verbosity / human-ness of the output would help it be less intimidating for new people
<gchristensen>
andreabedini: !!! I am so glad :D I spent a lot of time trying to make it really solid.
dejanr has quit [(Read error: No route to host)]
dejanr has joined #nixos
<andreabedini>
gchristensen: you did a good job then! it worked well
<andreabedini>
well, I'm not sure I fit the target audience :) if you say you're going to "use sudo a lot" I'll automatically f*** that
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bjornfor pushed 1 new commit to master: https://git.io/vQdR9
<NixOS_GitHub>
nixpkgs/master 8173649 Johannes Frankenau: neomutt: 20170609 -> 20170714
NixOS_GitHub has left #nixos []
<clever>
gchristensen: aha, even sourcing /etc/bashrc from ~/.bashrc doesnt work
<ToxicFrog>
Ok. Depending on how I use the version from unstable, either it doesn't run because it needs the kdewrapper when the rest of the system is on 17.03
<ToxicFrog>
Or it doesn't run because I'm trying to mix qt versions
<vaibhavsagar>
I tried mirror://pypi/b/bash_kernel/bash_kernel-0.6.tar.gz and that didn't work
<ToxicFrog>
I'm going to try just moving the entire system I want to run this on (which is fortunately not a general purpose system) to unstable and hope it works
<andreabedini>
vaibhavsagar: I think they have not uploaded any!
<vaibhavsagar>
not uploaded any?
<vaibhavsagar>
there's a wheel there
<vaibhavsagar>
do I just use that?
<andreabedini>
vaibhavsagar: yeah, since it's a pure package, a wheel it's just a tarball with the python files
newhoggy has quit [(Ping timeout: 248 seconds)]
<andreabedini>
vaibhavsagar: check yourselves tar tzf ~/Downloads/bash_kernel-0.6-py2.py3-none-any.whl
<andreabedini>
vaibhavsagar: the difference is only in the metadata
HurricaneHarry has joined #nixos
<vaibhavsagar>
oh, I see
<vaibhavsagar>
does Nix unpack wheels?
<andreabedini>
I remember reading that python wheels are kinda meant to be easy for re-distributors to repackage (so you make a .deb and a .rpm from a wheel)
<vaibhavsagar>
sure, but if I specify that in my src, does nix know what to do with it?
<vaibhavsagar>
let me try
<andreabedini>
one more question, can I install a specific version of a python package?
<andreabedini>
(my wet dream would be turn every requirements.txt into a nix configuration)
<gchristensen>
pypi2nix might be worth googling
<andreabedini>
gchristensen: awesome
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bjornfor opened pull request #27404: CONTRIBUTING.md: suggest "nixos/<module>" prefix for NixOS changes (master...contributing-add-nixos-prefix) https://git.io/vQd0b
NixOS_GitHub has left #nixos []
<ToxicFrog>
Ok it doesn't even build in unstable
* ToxicFrog
sobs
<ToxicFrog>
I just want a comic reader with ACBF support
<gchristensen>
ToxicFrog: :(
<ToxicFrog>
"Could not find a package configuration file provided by "Qt5LinguistTools""
<srhb>
Hmm. patchShebangs doesn't fix /usr/bin/env python -- is that intentional?
<joepie91>
quick question: how does NixOS, during boot, decide which init process to execute?
<joepie91>
as there's no /sbin/init, /bin/init, or /etc/init
<gchristensen>
joepie91: the path to init is in the bootloader
<gchristensen>
explore your /boot
<joepie91>
gchristensen: say that I'm rebuilding the system, how would I obtain the init path? the background is that I'm trying to get NixOS going under OpenVZ, and OpenVZ doesn't use initrd etc
<srhb>
Huh, the comments say it should.
<srhb>
Maybe I don't have python properly in.. scope or something
<andreabedini>
gchristensen: logging off, thanks!
<gchristensen>
andreabedini: you're welcome, good luck!
<ToxicFrog>
Hmm
andreabedini has quit [(Quit: leaving)]
<gchristensen>
joepie91: hmm I'd have to dig around, but maybe check out how netboot / pxe works -- see the nixos manual
<gchristensen>
joepie91: its outputted files might help
<ToxicFrog>
Is there somewhere that I can query the current status of each package in a channel?
<gchristensen>
joepie91: also did you see my PR? :)
<ToxicFrog>
See which ones are broken or not?
<joepie91>
will give that a shot, thanks
<joepie91>
gchristensen: the docs one? I didn't realize that was a PR
<ToxicFrog>
Because AFAICT peruse is just broken in unstable, full stop -- even if I remove the overrides it doesn't build.
<joepie91>
gchristensen: and well, I remain opposed to in-code docs and would like to explore alternative non-in-code docs options :)
<gchristensen>
joepie91: 5yeah go look :) gotta run tohugh
<gchristensen>
fair enough
<joepie91>
but I haven't had time to post in the thread yet
<joepie91>
gchristensen: I do hope it doesn't get merged before I can bring up said alternative options :P
<joepie91>
once it's merged, I expect the discussion to turn into "we have it now, make a compelling argument to get it removed"
newhoggy has joined #nixos
newhoggy has quit [(Ping timeout: 240 seconds)]
<gchristensen>
joepie91: unfortunately true :)
<gchristensen>
if things go according to plan, by the end of the day I'll have extensively improved the ruby docs. I hope you'll be able to review them.
* ToxicFrog
attempts a local checkout
raxius has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] rvl opened pull request #27405: postgresql: fix nixos tests and add xml support (master...postgresql-xml) https://git.io/vQdEQ
NixOS_GitHub has left #nixos []
watersoul has quit [(Ping timeout: 240 seconds)]
newhoggy has joined #nixos
<srhb>
Hmm. What are the differences between the environment of nix-shell --pure and nix-build?
<srhb>
Specifically, shell is finding /usr/bin/env python while build is not
<srhb>
Nevermind, I'm a doof.
newhoggy has quit [(Ping timeout: 240 seconds)]
Rotaerk has quit [(Ping timeout: 248 seconds)]
Rotaerk has joined #nixos
newhoggy has joined #nixos
<joepie91>
gchristensen: problem is that documentation, maintainability, etc. are things where small subtleties can make a big difference... and the moment the discussion turns into "prove that it should be removed" it becomes basically impossible to ever improve it again because every improvement in isolation looks insignificant and 'too much work'
<gchristensen>
I know
<joepie91>
and suddenly the same change that used to take a day of discussion now takes a month :P
<gchristensen>
sunk cost fallacy
<joepie91>
well yeah, but combined with the deceptive subtleness of improvements to non-code things
<gchristensen>
also yes
<joepie91>
it's very hard to convince people that a given improvement is going to make a difference
<joepie91>
it's much easier to have code thrown out :P
thblt has quit [(Ping timeout: 240 seconds)]
newhoggy has quit [(Ping timeout: 240 seconds)]
<vaibhavsagar>
hey, I have "ln -s ${thing}/bin/bash-notebook $out/bin/." in two different derivations, and it works in one but not the other
<joepie91>
gchristensen: I'd say that one big issue with that right now is that I can't easily identify 'bundlerEnv' in the text - I had to search around the text multiple times to figure out what it was talking about
<joepie91>
gchristensen: though I think that part of the problem is that this tries to be reference and guide at the same time
<gchristensen>
yes indeed
<gchristensen>
ok
newhoggy has joined #nixos
<joepie91>
gchristensen: also, idea: for function reference, it might be useful to reuse the frontend part of http://devdocs.io/
<joepie91>
which iirc is fairly separately usable from the rest of the stack
<gchristensen>
I'm going to finish what I'm planning, and ping you on the PR for critique
<gchristensen>
I think it'll be easier to suggest improvements from there
<joepie91>
alright
<joepie91>
thanks :)
<gchristensen>
I'll keep in mind to not try and mix them (but I don't anticipate much success >.>)
<joepie91>
gchristensen: I would suggest starting out with purely reference (modules, functions, arguments, return values, error conditions) in a structured format, and only writing the guides after
<joepie91>
means you have a frame of reference for what stuff to cover, and then you can just link to the reference from the guide where necessary
<joepie91>
gchristensen: also, I really strongly recommend playing around with devdocs a bit if you haven't used it before... it gives a pretty good idea of what a usable reference feels like
<joepie91>
UX-wise
ryanartecona has joined #nixos
<ToxicFrog>
Ok, I give up
<spinus>
vaibhavsagar: but what's the error?
<ToxicFrog>
We'll just have to settle for not having any metadata
<grahamc>
Awesome. Thank you a lot for the advice. I'll do that and take another crack at it after yard work :), joepie91
<ToxicFrog>
I tried doing a local nixpkgs checkout and editing the derivation there and it built fine and started up fine and then segfaulted when I tried to open anything, so I am just completely done
<joepie91>
grahamc: got super confused by your nick change for a moment :p
newhoggy has quit [(Ping timeout: 268 seconds)]
Wizek_ has quit [(Ping timeout: 260 seconds)]
Sonarpulse has joined #nixos
<simukis>
why is nixos putting all those drivers that aren’t necessary to mount root into the initrd?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] joachifm pushed 1 new commit to release-17.03: https://git.io/vQdgt
<NixOS_GitHub>
nixpkgs/release-17.03 a40677c Will Dietz: opensmtpd: Fix attempt to set unsupported permissions during install...
<simukis>
clever: thanks! I’ll try it. Exactly what I was looking for
<clever>
keep in mind, that mkForce will disable anything zfs was doing to help, so you need to manualy put it back in
<clever>
and you may also want to add usb back in
orivej has joined #nixos
<simukis>
clever: yeah, no problem
<simukis>
clever: another related question I had recently is why stage1 a shell spaghetti, instead of, you know, just using systemd target or something
<clever>
systemd isnt in the initrd
<simukis>
(could put the init system into initrd, is what I mean)
<clever>
that might make it more bloated
<simukis>
well, it is 1MB and it would avoid having to wait for all the devices to “udevadm settle” on boot.
newhoggy has joined #nixos
<clever>
systemd and its deps is 85mb
<simukis>
the init part of it is certainly not that large, but what do I know
* simukis
shrugs
<clever>
i avoided systemd like the plague before coming to nixos
<clever>
so i havent seen its initrd stuff
<bennofs>
Hmm, there also seems to be a thing called "dracut" for initrd
newhoggy has quit [(Ping timeout: 260 seconds)]
<Dezgeg>
it would be certainly nice to use it, but when I looked it was a massive pile of hardcoded /usr/bin/foo paths
vaibhavsagar_ has quit [(Ping timeout: 260 seconds)]
<Dezgeg>
(and even a bigger piece of shell scripts)
jgertm has joined #nixos
ryanartecona has quit [(Quit: ryanartecona)]
vaibhavsagar has quit [(Ping timeout: 276 seconds)]
newhoggy has joined #nixos
<S0rin>
Hey, I get an error when trying to compile master about mailutils anybody know how to fix it ? http://dpaste.com/13GFDHH
<Gravious>
S0rin, are you just compiling with -Werror ?
<S0rin>
Gravious: no just with nixos-rebuild
<Gravious>
yeah but is nix compiling with -Werror, maybe upstream isn't expecting -Werror ?
<S0rin>
Gravious maybe I will check but this seem's more related to wine than mailutils ?
<Gravious>
builder for ‘/nix/store/s51jqbgf1fmcyglywjbdbwzpsamgcdax-mailutils-2.2.drv’ failed with exit code 2
ibrahims has quit [(Ping timeout: 260 seconds)]
<S0rin>
Gravious, yes
newhoggy has quit [(Ping timeout: 260 seconds)]
Wizek_ has joined #nixos
<Gravious>
yet the warnings for mailutils seem to be all const qualifier related, which may be innocuous
<S0rin>
Gravious: yes
<bennofs>
S0rin: are there any other warnings before that?
ng0_ has quit [(Quit: Alexa, when is the end of world?)]
<bennofs>
normally, GCC tells you that it failed because of some -Werror=foo setting
<clever>
so: try building with "--option build-cores 1"
<clever>
S0rin:, oops
<Gravious>
bennofs, good point
<S0rin>
bennofs: did not find that kind of error
<bennofs>
S0rin: can you post the full log?
<S0rin>
clever: I am trying to rebuild with that option
<S0rin>
bennofs: yes
iyzsong has quit [(Ping timeout: 260 seconds)]
Ivanych has quit [(Ping timeout: 248 seconds)]
newhoggy has joined #nixos
<bennofs>
anyone here using neovim-ruby?
mrcheeks has quit [(Quit: ERC (IRC client for Emacs 26.0.50))]
jgertm has quit [(Ping timeout: 268 seconds)]
newhoggy has quit [(Ping timeout: 248 seconds)]
<jophish>
What's the simplest way of changing the libc version for a set of packages?
<jophish>
I'm certainly solving this problem in the wrong way though
<srhb>
jophish: Wrong is better than not!
<srhb>
(Imagine my code bases now, if you will)
<jophish>
hahaha
<S0rin>
I got some trouble to copy/paste it because it is bigger than my tmux session history but finally succeded here the full log http://dpaste.com/37RWBMP
<S0rin>
bennofs: here the log
<jophish>
I have this closed source binary (modelsim-ae) and it (obviously) doesn't have an RPATH which reads the nix store, so I've wrapped it with something setting LD_LIBRARY_PATH (I'm not patching the RPATH because it takes hours to install, and I want a quicker iteration time). If I put nix's glibc in the library path it just prints garbage and dies
<S0rin>
But basically the problem seem's to come from a problem on mailutils
<jophish>
So I have to let it use ubuntu's libc and friends
<srhb>
jophish: Sounds scary :P
<jophish>
except that nix wants 2.25, which ubuntu doesn't have yet!
<jophish>
I've no idea why I can't give it nix's glibc though
<srhb>
The hint isn't in the garbage?
<bennofs>
S0rin: ../config.h:1542:17: error: two or more data types in declaration specifiers
<bennofs>
#define ssize_t int
<bennofs>
S0rin: looks like the problem to me
<S0rin>
bennofs: yes so it don't come from the patches
ryanartecona has joined #nixos
newhoggy has joined #nixos
<S0rin>
And I try to rebuild without the patches in mailutils but the problem still occur
<S0rin>
is there a way to debug the problem ?
<clever>
jophish: steam has a similiar problem, it needs to use an older glibc, but then the mesa drivers on the host break with that older glibc
arianvp2 has quit [(Quit: arianvp2)]
newhoggy has quit [(Ping timeout: 246 seconds)]
newhoggy has joined #nixos
<S0rin>
bennofs: do you think this comes from my system state (I guess not) but as I understand it if it is broken in master it is broken for everybody who want to compile it
<bennofs>
S0rin: perhaps mailutils needs to be updated? The version in nixpkgs seems quite old (2.2), archlinux already has 3.2
newhoggy has joined #nixos
vcunat has joined #nixos
michas has joined #nixos
newhoggy has quit [(Ping timeout: 276 seconds)]
<jophish>
srhb: clever: I've solved my problem by not using any nix libraries :(
<jophish>
just installing the appropriate things on ubuntu
proteusguy has joined #nixos
<jophish>
ho hum
<clever>
:(
<jophish>
yeah
<srhb>
Where should asm-generic come from? It seems to exist in both linux and glibc
<srhb>
errno.h and errno-base.h specifically
<jophish>
I've got a replacement glibc build going, if that works I'll switch to that and let you know the good news, clever
<srhb>
Ah errno-base.h is only in linux, so I guess that's the one
<Sonarpulse>
vcunat: hi
<Sonarpulse>
do you think with configureFlags == null
newhoggy has joined #nixos
<Sonarpulse>
there should be no ---build --host --target?
<Sonarpulse>
before the logic was in the gcc cross wrapper
<Sonarpulse>
so null configureFlags would not prevent that
jtojnar has joined #nixos
<Sonarpulse>
jophish: thanks for the PR!
<jophish>
Sonarpulse: np, I sent you some messages on slack too
<Sonarpulse>
ah ok
<jophish>
nothing very important
<jophish>
Oh, Sonarpulse you'll be pleased to know that I got the SDRAM working this week
ryanartecona has quit [(Quit: ryanartecona)]
<Sonarpulse>
jophish: congrats!
<Sonarpulse>
I'm very curious what it took!
<jophish>
A bumch of qsys fiddling mainly
<jophish>
bunch*
newhoggy has quit [(Ping timeout: 268 seconds)]
<jophish>
the avalon drivers worked right away!
<Sonarpulse>
jophish: glad to hear it!
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vbgl pushed 1 new commit to master: https://git.io/vQdrg
<NixOS_GitHub>
nixpkgs/master 681f866 Vincent Laporte: qgis: 2.18.4 -> 2.18.10
NixOS_GitHub has left #nixos []
ixxie has joined #nixos
Infinisil has joined #nixos
newhoggy has joined #nixos
proteusguy has quit [(Ping timeout: 276 seconds)]
newhoggy has quit [(Ping timeout: 240 seconds)]
<jophish>
clever: that worked (the older glibc). srhb it was as easy as just overriding src (and patches = [])
<jophish>
thanks!
newhoggy has joined #nixos
<srhb>
jophish: whee. :)
<srhb>
Despite having trouble with complicated builds still, I still appreciate Nix so damn much.
<srhb>
<3
Wizek has joined #nixos
<jophish>
srhb: I can't stand not to use it
<jophish>
even for complex builds, once it's done once it's done
<jophish>
(forever)
<clever>
until somebody upstream changes something
<srhb>
clever: There is no change, ever! >:/
<srhb>
(except magically changing versioned files. Looking at you Mozilla.)
newhoggy has quit [(Ping timeout: 240 seconds)]
<simukis>
clever: I managed to make my initrd 1MB leaner by jetissoning all those unnecessary modules
praduca has joined #nixos
<gchristensen>
was 1MB worth the effort?
<gchristensen>
(genuine question)
<clever>
simukis: nice
<praduca>
Hi, anyone knows how I can download nix expressions for haskell packages? they don't appear to be on nixpkgs
<simukis>
gchristensen: maybe not… but I was also dealing with other stuff at the same time, like figuring out how to limit ARC size (you can’t do that modprobe config because module is loaded before modprobe configs are available)
<clever>
simukis: kernel commandline
<simukis>
yeah
<gchristensen>
praduca: they're in haskellPackages: nix-env -qaP -A nixpkgs.haskellPackages or nix-env -qaP -A nixos.haskellPackages
<gchristensen>
simukis: ok, cool, I got it :)
<clever>
42 kernelParams = [
<clever>
47 "zfs.zio_taskq_batch_pct=50"
<clever>
48 "spl.spl_taskq_thread_bind=1"
michas has quit [(Quit: Leaving)]
<clever>
simukis: this method even works for modules that have been compiled into the kernel
[0x4A6F] has quit [(Ping timeout: 260 seconds)]
<simukis>
in general I believe that nixos could do a lot early boot optimisation still. It annoys me that every boot I have to wait for all the bazillion spinning rusts to spin up before it does anything else, even though the OS is whole in a SSD attached via m.2
<gchristensen>
that is reasonable :)
<aristid>
simukis: you could bypass the fileSystems option, and make your own systemd units for mounting
<simukis>
aristid: I only mount the root filesystem and /tmp with fileSystems
<simukis>
not even boot
newhoggy has joined #nixos
portu has left #nixos ["ERC (IRC client for Emacs 25.1.1)"]
schoppenhauer has joined #nixos
<simukis>
oh, boot too, because nixos-rebuild complains without it
<simukis>
IIRC
<aristid>
simukis: so why does it wait for them to mount?
<clever>
if /boot isnt mounted at rebuild time, the system will just rollback all changes every time you reboot
<clever>
and if you GC, it can cease to boot entirely
<simukis>
aristid: it doesn’t wait for them to mount, it waits for all the disks (AFAICT at least) to `udev settle`
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Ericson2314 opened pull request #27407: mkDerivation: Fix errors from #27365 (staging...null-configure-flags) https://git.io/vQdKr
NixOS_GitHub has left #nixos []
<aristid>
simukis: ah
<simukis>
aristid: which involves spinning them all up and doing some initialisation for all of them
<simukis>
clever: I believe in auto-mount
<aristid>
simukis: clearly the solution is an nvm-e only system! :D
<simukis>
:P
vcunat has left #nixos []
<simukis>
no way anybody needs to have /boot mounted most of the time, past UEFI getting the two `.efi`s off it and running them
<simukis>
(also yeah, upgrades)
<schoppenhauer>
hi. is it possible to define a (local) package with a git repo that is not public? like, specifying an SSH-Key or something?
<clever>
aristid: too many machines that lack nvme support in the uefi firmware
<aristid>
schoppenhauer: sure, the easiest way is probably to just add it to your package overrides in the nixpkgs config
<schoppenhauer>
aristid: but how do I specify it?
<Sonarpulse>
jophish: glibc problem
<Sonarpulse>
?
<gchristensen>
schoppenhauer: one possibility is to use fetchgitPrivate, but much more common is to use a local clone and put it in your NIX_PATH
<Sonarpulse>
because on my cross-elegant branch, I happen to have one
newhoggy has quit [(Ping timeout: 248 seconds)]
<aristid>
clever: if you can afford nvm-e only storage, you can probably also afford a new motherboard :)
<aristid>
anyways, maybe nvm-e is not cost effective but compared to all my sata ssds, my new nvm-e ssd is amazingly fast
<simukis>
aristid: but can you notice the difference
<gchristensen>
325,000 read IOPS over pcie
<simukis>
also amazingly hot
<aristid>
simukis: i can format the whole damn drive in a few seconds.
<gchristensen>
and 6TB per card
<simukis>
aristid: I can clear the whole damn drive with 0s in a second
<simukis>
(cough trim cough)
<aristid>
simukis: i meant with a secure erase :)
<schoppenhauer>
clever: is it possible to specify the key/password/etc. in the expression?
<aristid>
also nvm-e enables things that the low-level fanboy in me likes: the idea of userspace storage just seems cool to me ( http://www.spdk.io/ )
<clever>
schoppenhauer: any attempt to do that will result in the key landing in /nix/store and being world-readable
<simukis>
> operating in a polled mode instead of relying on interrupts
* simukis
raises an eyebrow or three
<schoppenhauer>
clever: ok. hm. I am not sure I understand which users must have the right to do ssh-access. do I need to give all nixbld-accounts read-access?
<aristid>
simukis: unfortunately that's the only way to get the best latency :)
<clever>
schoppenhauer: the ssh agent socket handles that, so the build users never need read access
<aristid>
simukis: but yeah in practice this approach is too extreme in 99.9% of cases
<simukis>
aristid: the only way to get the best latency is to operate in ring0
<aristid>
simukis: i just find it cool, and find it a pity i have no reason to do it :D
* simukis
shrugs
<praduca>
gchristensen: Hi, thanks. There is a way to download the nix expression for one particular package? the regular ones are on the git repo, but i cant find the ones for the haskell packages are...
<simukis>
I guess if you’re in the userspace, polling avoids going through the OS, which is maybe nice
<clever>
praduca: they are in a file called hackage-packages.nix, its massive
<schoppenhauer>
clever: my actual problem comes from me trying to create test systems for some private code using nixOps. so I do not really know whether that ssh-agent stuff will work.
<praduca>
clever: thanks!
<clever>
schoppenhauer: i cant see any reason why the agent wouldnt work, just start an agent and run ssh-add on a key that has access
<schoppenhauer>
clever: isn't a part of the stuff built on the virtual machines?
<clever>
schoppenhauer: are you on a mac?
<schoppenhauer>
clever: no. nixos.
<clever>
it doesnt use virtual machines on nixos
<aristid>
but it does use the nix-daemon
<schoppenhauer>
ok.
jtojnar has quit [(Quit: jtojnar)]
<aristid>
not sure which user it runs the fetchgit as
<clever>
a member of the nixbld group
<schoppenhauer>
ok, maybe a simpler method would be to just have a local clone … does fetchgit work with local clones too?
<clever>
there is fetchgitlocal for that
<clever>
pkgs.fetchgitLocal
* simukis
notes fetchgitlocal, because it is super relevant to them later on
<aristid>
you can also just do ./path/to/local/repo
<clever>
yeah, and just skip the git layer entirel
<clever>
y
<aristid>
although that does import the whole thing into the store, with files you might not need
<johnw>
hmm.. this is strange, when I cp -pR ${src2}, the resulting directory has some of the files I need, but not all. However, there's a strange subdirectory in there named ayrrgcix65h09dxn1bfga0x2zvv0j8hb-CompCert-2e2a266 which does have exactly the contents I want
<clever>
the directory gets renamed
<clever>
./foo winds up at /nix/store/<hash>-foo/
<johnw>
this is a local directory to the that directory
<clever>
and if you dont tell cp what the destination name is, it keeps the hash
<bennofs>
gchristensen: is it really necessary to call every package that's build from git commit -unstable?
<bennofs>
gchristensen: that seems a bit too much. I can understand it if there's two versions of the package in the repo
<gchristensen>
I'm not sure
<bennofs>
gchristensen: just wondering, for example for dosbox. the latest release was in 2014, so should we call it dosbox-unstable-date or just dosbox-date ?
<gchristensen>
well unstable-date b/c it is unreleased software
slack1256 has joined #nixos
<bennofs>
gchristensen: hmm. but that means that nix-env -i dosbox will fail, right? wouldn't dosbox-date-unstable be better? do you know if there was some discussion about this choice somewhere already?
<gchristensen>
I'm not sure I've ever used nix-env -i :/ I'm the wrong person to be asking about his :P
<bennofs>
i don't use it either :)
newhoggy has joined #nixos
Winchell1M has joined #nixos
<Winchell1M>
Does anybody have any advice for how to set up a wpa wireless network connection in configuration.nix?
<Winchell1M>
@srhb: Yes I think you're right I just looked it up.
<Winchell1M>
Do you know if this will automatically connect when I `nixos-rebuild switch` and/or reboot?
<Winchell1M>
Or do you still have to manually connect somehow?
newhoggy has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Infinisil opened pull request #27408: bashSnippets: init at 1.12.0 (master...bashSnippets) https://git.io/vQdPC
NixOS_GitHub has left #nixos []
<srhb>
Winchell1M: I think if networking.wireless.enable = true; it will auto connect
<clever>
and if wireless.enable = false;, it wont even install wpa_supplicant, so you have no way to manualy start it
newhoggy has quit [(Ping timeout: 268 seconds)]
<srhb>
Simple!
<Winchell1M>
Ugh
<Winchell1M>
I can't nixos-rebuild
<Winchell1M>
To get wifi working
<Winchell1M>
Because I'm not connected to wifi
<clever>
yeah, common problem, you either need to boot from the installer and use "nixos-install --chroot" or plug in an ethernet cord
<gchristensen>
joepie91: it is super hard to do what you're suggesting
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nix] alicebob opened pull request #1467: use sha256 hashes in the examples (master...nomd5) https://git.io/vQdPF
NixOS_GitHub has left #nixos []
<Winchell1M>
@clever: Do you mean boot from USB and literally run that command?
<Winchell1M>
Or does it take arguments
<clever>
Winchell1M: after mounting everything under /mnt like you did for the install, run the above command to get a shell inside that /mnt
<clever>
Winchell1M: then you can "nixos-rebuild boot" to update the system
ibrahims has quit [(Ping timeout: 240 seconds)]
\\\\\\\\\\\ has joined #nixos
newhoggy has joined #nixos
<Winchell1M>
Ok
vaibhavsagar has joined #nixos
newhoggy has quit [(Ping timeout: 260 seconds)]
<Winchell1M>
@clever: I followed those instructions
<Winchell1M>
Now I get "Parent directory (/run/user/0) does not exist at /nix/store/...
<srhb>
Oh boy, rebuilding ceph 20 times over and progressing a few percent each time is so much fun.
newhoggy has joined #nixos
<bennofs>
srhb: nix-shell ?
ibrahims has joined #nixos
<srhb>
Winchell1M, clever: Another user had that error earlier today as well. I don't think running nixos-rebuild from within the chroot works anymore
<srhb>
bennofs: I am, but a lot of the changes are reconfiguring the cmake generating script at the very start, which for some obscure reason needs to nuke the build directory.
<srhb>
bennofs: !!!FUN!!!
<bennofs>
:/
<srhb>
I could probably do a bit better if I dug into it, but, cba.. :-P Almost done, judging by the percentages anyway.
ixxie has quit [(Quit: Lost terminal)]
newhoggy has quit [(Ping timeout: 240 seconds)]
<bennofs>
wtf. does anyone know how it can happen that `dig` times out if I dig @172.0.0.1 packages.gentoo.org but doesn't if I dig @::1 packages.gentoo.org ??
<bennofs>
this is dnsmasq running locally
[0x4A6F] has joined #nixos
ryanartecona has joined #nixos
<Winchell1M>
@srhb: If I don't have an ethernet port, do you have any ideas for what else I could try?
<srhb>
Winchell1M: Sorry, why did you need an ethernet port again?
[0x4A6F] has quit [(Remote host closed the connection)]
<Winchell1M>
I'm trying to get wifi working on a fresh nixos install
[0x4A6F] has joined #nixos
<Winchell1M>
I edited my configuration.nix to do so
newhoggy has joined #nixos
<srhb>
Winchell1M: You can just rerun the install process with the fixed configuration.nix with wifi enabled (minus the formatting)
<Winchell1M>
And when I run nixos-rebuild switch it needs internet to do so
<Winchell1M>
Ah -- yes that should do it
newhoggy has quit [(Ping timeout: 260 seconds)]
ryanartecona has quit [(Quit: ryanartecona)]
newhoggy has joined #nixos
Itkovian has joined #nixos
newhoggy has quit [(Ping timeout: 240 seconds)]
newhoggy has joined #nixos
arianvp2 has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Ericson2314 pushed 2 new commits to staging: https://git.io/vQdMu
<NixOS_GitHub>
nixpkgs/staging aaaa470 John Ericson: mkDerivation: Fix errors from #27365...
<NixOS_GitHub>
nixpkgs/staging 7d95391 John Ericson: Merge pull request #27407 from obsidiansystems/null-configure-flags...
NixOS_GitHub has left #nixos []
newhoggy has quit [(Ping timeout: 240 seconds)]
michaelpj has quit [(Ping timeout: 260 seconds)]
<gchristensen>
guh for all I don't love it, you can usually guess how to write docbook by just thinking about what you're writing. <function>bundlerEnv</function> <parameter>gemConfig</parameter> <emphasis>extend</emphasis>
DutchWolfie has quit [(Quit: Konversation terminated!)]
<jophish>
How can I check given nixpkgs if it's a cross build set?
<jophish>
Sonarpulse:
<jophish>
or better yet, some haskell.packages.foo
takle has quit [(Remote host closed the connection)]
vaibhavsagar has quit [(Ping timeout: 240 seconds)]
takle has joined #nixos
<gchristensen>
cannot coerce a set to a string, at /nix/store/hq64pzjfjk18gh31wz47lvgghi2mp8q0-nixos-17.03.1517.849b5a5193/nixos/pkgs/build-support/fetchgitlocal/default.nix:4:12
<gchristensen>
this is the important bit
pie_ has quit [(Ping timeout: 240 seconds)]
newhoggy has joined #nixos
<gchristensen>
hm
<gchristensen>
schoppenhauer: the problem is you're passing a set, it just needs a string. so: src = pkgs.fetchgitLocal { src = "/home/christoph/tmp/nixops/chat"; };
<gchristensen>
but you need to do: src = pkgs.fetchgitLocal "/home/christoph/tmp/nixops/chat";
tmaekawa has quit [(Ping timeout: 248 seconds)]
<srhb>
Holy crap, it built. Now I get to do it all over with nix-build >>
<wizwizwizwiz>
hi i'm trying to use nix as a package manager on ubuntu 16.40 to try reflex (ghcjs)
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<wizwizwizwiz>
*ubuntu 16.04
<srhb>
wizwizwizwiz: Yes?
arianvp2 has quit [(Remote host closed the connection)]
<wizwizwizwiz>
i've had a few problems with missing symbolic links in /usr/lib/x86_64-linux-gnu but i managed to overcome that
<wizwizwizwiz>
and a couple of missing dev packages
arianvp2 has joined #nixos
<srhb>
That sounds peculiar.
<srhb>
You're using the tryreflex repository?
<wizwizwizwiz>
yeah
<gchristensen>
schoppenhauer: so now the problem is the builder can't access "/home/christoph/tmp/nixops/chat" because it doesn't have arbitrary access to your filesystem, you're passing a string, you should pass it something of the datatype path.
<wizwizwizwiz>
and it complains that i have an out of date nix
<wizwizwizwiz>
so when i run nix-env --upgrade
Ivanych has joined #nixos
<wizwizwizwiz>
i get this error:
bennofs has quit [(Quit: WeeChat 1.9)]
takle has quit [(Ping timeout: 260 seconds)]
<wizwizwizwiz>
error: couldn't change to directory of ‘/nix/var/nix/daemon-socket/socket’: No such file or directory
<gchristensen>
schoppenhauer: ie: src = pkgs.fetchgitLocal ./chat; -- assuming this is happening in /home/christoph/tmp/nixops
newhoggy has quit [(Ping timeout: 246 seconds)]
<srhb>
wizwizwizwiz: How did you install nix?
<wizwizwizwiz>
i do have a /nix/var/nix folder but there's no daemon-socket
<wizwizwizwiz>
to be honest i forgot how i installed nix
<wizwizwizwiz>
when i run nix-env --version it complains that the nix database directory isn't writable so i can't get the version
<srhb>
wizwizwizwiz: Nuke it from orbit and start again, taking note of the steps.
<wizwizwizwiz>
oh right... when i ran the try-reflex install script
<manveru>
gchristensen: how did it go with your docs? :)
<wizwizwizwiz>
it was complaining about missing libraries etc and it would barf
<gchristensen>
manveru: some people like it, joepie91 don't :)
szn has quit [(Quit: WeeChat 1.4)]
<gchristensen>
I'm going to head out for a while.
<gchristensen>
been writing new docs all day ...
<wizwizwizwiz>
and every time i fixed a library it would re-download a big shell script
<wizwizwizwiz>
so i downloaded the script myself
<wizwizwizwiz>
so i could re-run it rapidly as i fixed libs
newhoggy has joined #nixos
<wizwizwizwiz>
how do i "nuke nix from orbit" ?
<gchristensen>
wizwizwizwiz: what platform are you on?
<wizwizwizwiz>
is it enough to rm -rf /nix ??
<wizwizwizwiz>
whoops meant to type one question mark there. i'm on ubuntu 16.04
<wizwizwizwiz>
i'm super noob when it comes to nix (and haskell, etc)
<wizwizwizwiz>
like double clicking, how to buy cheap printer ink, etc
<manveru>
i'll take a look, thanks
<srhb>
wizwizwizwiz: I think that should be sufficient to get started. Then, I would suggest trying to install nix from the regular source rather than through try-reflex, just in case.
<srhb>
wizwizwizwiz: By the way, in case you didn't know, there's a #reflex-frp channel on Freenode as well, and they are really helpful there, too. (Your questions are fine here too though)
<schoppenhauer>
manveru: and using bundlerenv would mean that I would have to write a gemfile for this.
<manveru>
does it have a gemspec?
<schoppenhauer>
no
newhoggy has joined #nixos
<schoppenhauer>
it's written by someone else in my team …
<manveru>
what program is it?
<schoppenhauer>
private stuff
<schoppenhauer>
not public
<manveru>
well, on debian all the gems packaged as deb files are put into one directory basically
<manveru>
depending on what dependencies there are, you can try using RUBYLIB env var to point to the lib directory of each
<manveru>
i can't really help you much more, it's pretty late and i'm not a big fan of blind debugging :P
newhoggy has quit [(Ping timeout: 260 seconds)]
<schoppenhauer>
manveru: ok, RUBYLIB seems to work
<schoppenhauer>
manveru: thxy
newhoggy has joined #nixos
<Infinisil>
wizwizwizwiz: Are you wizzup on github?
wak-work has joined #nixos
<manveru>
np
newhoggy has quit [(Ping timeout: 260 seconds)]
leat has quit [(Ping timeout: 240 seconds)]
hiratara has quit [(Ping timeout: 255 seconds)]
<srhb>
Note to self, don't install Ubuntu desktop on just one core....
Infinisil has quit [(Ping timeout: 248 seconds)]
justbeingglad has joined #nixos
hiratara has joined #nixos
<wizwizwizwiz>
ok so after that amount of time ghcjs-boot: error running: /home/u/.stack/snapshots/x86_64-linux/lts-3.6/7.10.2/bin/cabal --config-file /home/u/.ghcjs/x86_64-linux-0.2.0-7.10.2/ghcjs/cabalBootConfig --ignore-sandbox install ./boot/cabal/Cabal --global --ghcjs --one-shot --avoid-reinstalls --builddir dist --with-compiler /home/u/.stack/programs/x86_64-linux/ghcjs-0.2.0.20160202_ghc-7.10.2/src/.stack-work/install/x86_64-l
<srhb>
wizwizwizwiz: Try in #stack or #reflex-frp or even #haskell for that.
Infinisil has quit [(Client Quit)]
newhoggy has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] volth opened pull request #27412: geoip: set default dir to '/var/lib/geoip-databases' (master...geoip-default-dir) https://git.io/vQdH9
NixOS_GitHub has left #nixos []
bennofs has quit [(Ping timeout: 240 seconds)]
python476 has quit [(Ping timeout: 260 seconds)]
newhoggy has quit [(Ping timeout: 260 seconds)]
orivej has quit [(Ping timeout: 248 seconds)]
newhoggy has joined #nixos
orivej has joined #nixos
<gchristensen>
the -> operator always makes me confused, and requires I go try the whole truth table in nix-repl ...
<gchristensen>
but it is a nice operator. any tips on remembering its use?
<slack1256>
implication?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to staging: https://git.io/vQdQn
<NixOS_GitHub>
nixpkgs/staging c3d9ec5 Tuomas Tynkkynen: pythonPackages.unittest2: Fix missing argument to substituteInPlace
NixOS_GitHub has left #nixos []
<Dezgeg>
yeah, isn't it the standard mathematical logic operator
<slack1256>
I don't know if all colleges pass some logic before calculus. I can be standard, yet not usually known
<gchristensen>
yeah it is, but still. my technical upbringing wasn't really academic
<gchristensen>
I guess I just need to practice :)
ryanartecona has joined #nixos
newhoggy has quit [(Ping timeout: 260 seconds)]
<jophish>
you could try and remember that `a -> b` is `!a || b`, then there's no need for the truth table
<gchristensen>
hmm yes that might be helpful
* tilpner
prefers a && !b
<jophish>
the implication is clearer with !a || b (for me at least :)). If a is true, then b must be true, if a is false then there is no constraint on b
<jophish>
actually, it's pretty similar with a&&!b
<slack1256>
by de morgan laws
<gchristensen>
tilpner: that is much easier for me to grok for some reason
Infinisil has joined #nixos
<slack1256>
tilpner: should (a -> b) == !(a && !b) ?
<tilpner>
Yeah, hold on, I think I'm wrong
<gchristensen>
yeah I'm not sure that works
<gchristensen>
anywho, I'll just try applying more rigor :P
<tilpner>
Sorry. "a -> b is false when a is true but b is false. true in all other cases" is what I remember about it, not sure how that lead to a && !b
<Infinisil>
gchristensen: How is the docstring-type-of-thing going?
jsgrant has quit [(Read error: Connection reset by peer)]
<slack1256>
as !(a && !b) == (!a || b) == (a -> b)
<tilpner>
And !(a && !b) is of course not any easier to reason with than !a || b
<simpson>
Yep. `a -> b` is false if a does not, in fact, imply b, and otherwise true. That's how I remember it too.
<slack1256>
first step by de morgan, second by material implication
<slack1256>
basically in implication you only have to consider the case when the antecendent is true.
<gchristensen>
Infinisil: pretty good, but moretea has a moreinteresting idea
<slack1256>
it is then obliged the consequence to also be true.
<slack1256>
That is the only law
<Infinisil>
gchristensen: Heh, who is moretea? What's the idea?
<simpson>
Huh, that's interesting. I never thought about defining exponentiation in Boolean algebra.
Rotaerk_ has quit [(Ping timeout: 260 seconds)]
<Infinisil>
jophish: But that doesn't work well for a=b=0
Rotaerk has quit [(Ping timeout: 248 seconds)]
pie_ has quit [(Ping timeout: 268 seconds)]
<jophish>
Infinisil: hehe, I'm defining 0^0 = 1
<simpson>
Infinisil: The elements in Boolean algebra are {true, false}, not {0, 1}. Addition is OR; multiplication is AND.
<Infinisil>
gchristensen: Well I think our idea is more interesting, but his sticks to a more stringbased approach and therefore doesn't need big refactoring
<Infinisil>
simpson: If we'd look at it like this then b^a would be XOR, which is probably not what jophish meant :)
<simpson>
Infinisil: Under a certain extremely natural and incredibly non-intuitive definition of how numbers should work, 0 ** 0 should be 1, and thus the argument of `false ** false == true` is not so silly.
newhoggy has joined #nixos
<Infinisil>
simpson: Well I don't wanna start a discussion on how to define 0^0, in the end it would make *some* sense both ways
<jophish>
Knuth says that 0^0=1 is a definition and not a fundamental mathematical truth (whatever that means)
pie_ has joined #nixos
<simpson>
Infinisil, jophish: There's two solid justifications that are easy to understand. First, there's the discrete argument: The Binomial Theorem needs special cases if you define it to anything else.
<Infinisil>
jophish: Basically this: mathematical axioms are nothing more than definitions
newhoggy has quit [(Ping timeout: 240 seconds)]
<slack1256>
definitions that don't contradict each other*
<simpson>
The continuous argument: Pick analytic functions f and g, make sure they're defined and that f stays positive, and then take the limit f(x) ** g(x) as x -> 0. The limit goes to 1.
<Infinisil>
We can define anything we want, but we want to define things so that interesting and practical stuff turns up
<jophish>
Infinisil: exactly :)
<simpson>
Infinisil, jophish: Maybe. It's also totally possible that maths is discovered, not invented. I wouldn't be so sure of either position.
<jophish>
I've always thought that "Math is discovering properties of things we invent"
<Infinisil>
We're shifting from #nixos -> mathematics -> philosophy
<jophish>
help me! I've got segfaults in my setups: /nix/store/zcp0m09vhwzv1azfj28lvbkcm525bbhn-stdenv/setup: line 80: 22904 Segmentation fault
<simpson>
jophish: Maybe we invented the primes? But isn't that terrifying?
<jophish>
simpson: we should have invented them to be easier to reason about then!
<dash>
jophish: i feel that way about most programming languages
<jophish>
hahahaha
<simpson>
jophish: Easier said than possible.
* Infinisil
doesn't get it
<slack1256>
this is the only philosophy discution that follow any math problem
elninja44 has joined #nixos
<simpson>
slack1256: What, philosophy of maths is related to maths? No way!
<slack1256>
the only question philosophy of maths ask is if math is discovered or invented. There is nothing else that I know
<catern>
simpson: what would anthropic bias in primes mean?
<slack1256>
other questions are usually logic problems and the foundations programs claim dibs on them
<simpson>
catern: It'd suggest that maths doesn't reflect reality, which is surprising since it's the theory of knowledge we have which is closest to reality.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] volth opened pull request #27413: varnish service: made compatible with varnish 5.2.1, add modules (master...varnish-fixes) https://git.io/vQd7M
NixOS_GitHub has left #nixos []
<simpson>
slack1256: Ah, you've never met a finitist, let alone an ultrafinitist. There are some truly surprising POVs out there.
<dash>
let's create a non-standard axiomatization of math based on the actor model
<catern>
simpson: no, I mean, what would a world with anthropic bias in primes even look like?
<simpson>
dash: Pi calculi are Turing-complete.
<Infinisil>
simpson: And that's exactly the point of philosophy, there are always different viewpoints
<catern>
simpson: i guess - what do you even mean by "anthropic bias"?
<srhb>
wizwizwizwiz: The nix installer works almost out of the box on 16.04, I just needed to install curl
<simpson>
catern: I don't know. Literally everything that you and I ever observe is anthropically biased.
<dash>
simpson: alternate hypothesis: we live in an anthropically biased universe
<simpson>
catern: The anthropic bias is an inherent and unconscious bias in humans. It causes all humans to percieve reality through human senses and interpret reality using human thoughts.
<Infinisil>
srhb: Ubuntu without curl? Huh?
<simpson>
dash: The anthropic principle!
<srhb>
Infinisil: Maybe a quirk of the desktop version?
<catern>
simpson: i mean, yes, i'm familiar with it
hyphon81_ has joined #nixos
<simpson>
Infinisil: Some builds of some distros come without either curl or wget, on the principle that anybody who knows how to use one can usually use the other.
<Infinisil>
I'll check the start of my bash history to see if i needed to install anything before I was able to install nix
<simpson>
catern: So, what I'm saying is: What if numbers as they really exist aren't something we perceive, but instead we only perceive their shadows? What if the ring of integers, and their ideals the primes, are actually just quirks of the way that our minds work? There's no compelling reason for this not to be the case other than that it's always surprising to be reminded that humans don't directly observe reality.
<catern>
eh, i don't know that "anthropic bias" is the best word to use for that...
pie_ has quit [(Read error: Connection reset by peer)]
<catern>
i mean, the primes don't seem to be a necessary condition for intelligence to exist
<Infinisil>
Nope, I had curl on my Ubuntu, using the 17 version though
<slack1256>
simpson: in kantian terms, are you asking if numbers are analityc or syntetic ?
pie_ has joined #nixos
pie__ has joined #nixos
<catern>
generally when I think of the anthropic principle, I think about the fact that, for example, we observe an ordered physical universe
<catern>
but in fact, if there was on ordered physical universe, there would be no humans
<simpson>
slack1256: Sure.
<catern>
you might say the same of math in general, but picking out the primes in particular - seems unrelated to the anthropic principle
<catern>
if there was no ordered physical universe*
<simpson>
catern: I am trying to introduce you to something new, then; do you think that that's air that you're breathing? Why do you think that your *observations* reflect the actual *nature* of reality?
pie_ has quit [(Remote host closed the connection)]
<catern>
rest assured I am very familiar with this concept
<catern>
i just don't think it's quite the same thing as the anthropic principle
<catern>
it's just general, you know, "all experience is subjective", stuff
<simpson>
The anthropic bias is really tightly related, though. After all, the anthropic principle is really just boasting that the universe seems, when *we* observe it, to follow rules that make *us* possible. Doesn't that seem like, well, a biased POV?
\\\\\\\\\\\ has quit [(Quit: leaving)]
<catern>
the anthropic principle/anthropic bias is a specific phenomenon, related to the fact that our experience draws from a sample where we can experience
<catern>
yes, it's certainly biased
elninja44 has quit [(Remote host closed the connection)]
rtjure has joined #nixos
<simpson>
catern: Anyway, I'll drop this, since we've totally crowded out srhb. Sorry.
<catern>
right, indeed
<dash>
"recent measurements with more accurate instruments have fixed the value of the first natural number at 1.000000043"
<simpson>
dash: Have you been following the redefinition of the kilogram? It's literally that.
<dash>
pretty much
newhoggy has joined #nixos
<Infinisil>
I wonder if america/englang will ever adapt the metric system
<simpson>
Infinisil: Last time I was in the UK, it was 100% metric. The USA was one of the first signatories to metric and all of the old units are actually defined based on metric units.
<srhb>
simpson: What, people weights finally shifted?
ibrahims has quit [(Ping timeout: 248 seconds)]
newhoggy has quit [(Ping timeout: 240 seconds)]
<simpson>
srhb: Sure? I know both. I suspect that another generation or two will be needed to finish the transition, and some niches may never change due to traditions.
<simpson>
In particular, there's only one thing I can think of in the USA measured in slugs.
<pstn>
I want to use exactly the same openssh configuration in an declarative container as in my host. How would I achieve that?
akaWolf has quit [(Ping timeout: 260 seconds)]
akaWolf has joined #nixos
<simpson>
pstn: Use the same version of nixpkgs and nixos in both, and then use the same services.openssh clauses in both. That should be sufficient, I think? There's a couple things like stateVersion that affect SSH which you should keep identical too.
<Infinisil>
pstn: Create a file with `{ config, pkgs, ...}: { <your openssh config> }` then import that file from both configurations
<Infinisil>
(and what he said)
<dash>
Infinisil: we're never going to adopt metric here
<Infinisil>
dash: $5 it gets adopted in the us, eventually
<dash>
Infinisil: I doubt the USA will last that long
<Infinisil>
Heh
<dash>
Infinisil: anyway, they tried to put up kilometer markers in the highways here in the 90s, in addition to mile markers
<dash>
Infinisil: some folks went around knocking them all down
<Infinisil>
Ohhh..
ng0 has joined #nixos
<simpson>
Infinisil: It's already adopted by every part of the USA except the cultural fabric. Y'know, that part that can't be rushed.
<Infinisil>
Well I guess I should've expected that from MURICA
<dash>
Infinisil: it was an entirely reasonable thing to do IMO
<Infinisil>
simpson: Adopted by every part? I have never ever seen americans use centimeters (in videos that is)
<dash>
Infinisil: it's used where it's useful, engineering and physics
<simpson>
Infinisil: And yet, we're taught in schools that 1in == 2.54cm (accurate to nearly five places IIRC), and we know how to convert. That's what I'm trying to say: We know *both* sets of units.
<Infinisil>
dash: Ah yes, forgot about that
<dash>
Infinisil: anything else is the work of the devil
<Infinisil>
simpson: Well that's nice
<pstn>
Thanks for your tipps, this should make my config easier to work with in the future.
<simpson>
We also know that 2.2lb ~ 1kg (close enough, right?), so I can boast that I recently got below triple-digit kg.
<Infinisil>
simpson: Congrats!
justelex_ has joined #nixos
<Infinisil>
I have this problem where the external display only works when I have it plugged in during startup, but if i dis/reconnect it doesn't anymore (audio to the display still works though)
<Infinisil>
Maybe there is some systemd service that does a thing at startup, any idea how to find out what's the problem?
zeus_ has joined #nixos
<spinus>
does xrandr or some display configurator see that?
<Infinisil>
xrandr doesn't see the display anymore after reconnecting
zeus_ has quit [(Remote host closed the connection)]
<spinus>
I would check startup logs and try to find device init, maybe some module is inserted and then removed?
zeus_ has joined #nixos
<Infinisil>
It's connected via displayPort, (it's an apple thunderbolt display), and is using the usbhid driver (`usb-devices` shows that)
newhoggy has joined #nixos
<spinus>
do you see usb stuff in logs? does device exist in lsusb output?
<Infinisil>
spinus: Yup it exists in lsusb and usb-devices
<spinus>
(btw, displayport works for me)
<spinus>
so maybe x cannot detect the change
<Infinisil>
I can see kernel messages about it discovering the display device
<Infinisil>
I already tried restarting the display-manager, but to no avail
<spinus>
ah, that's interesting
<Infinisil>
Are there any well knows commands to work with pci device / usb / hid stuff?
<Infinisil>
There is udevadm I guess, I'll play around with it a bit
<spinus>
Infinisil: one thing you could do is to do snapshot of some settings when display works and when it does not and try to compare
<spinus>
for example "find /dev" "find /sys"
<spinus>
maybe that way you can find "something" which will lead you to the source
<Infinisil>
spinus: I did that with lsusb and usb-devices, but had some problems with seeing the diffs heh
<Infinisil>
spinus: I'll try again with "find /dev", thanks
markus1199 has joined #nixos
Wizek has quit [(Ping timeout: 240 seconds)]
erictapen has joined #nixos
Wizek_ has quit [(Ping timeout: 260 seconds)]
<spinus>
Infinisil: when display works and you are turning is off, check if any module is not removed
<spinus>
udevadm monitor could help
hyphon81_ has quit [(Remote host closed the connection)]
markus1189 has quit [(Ping timeout: 240 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] NeQuissimus pushed 5 new commits to master: https://git.io/vQdFJ
<NixOS_GitHub>
nixpkgs/master c5ef98b Tim Steinbach: linux: 4.12.1 -> 4.12.2
<NixOS_GitHub>
nixpkgs/master ccec165 Tim Steinbach: linux: 4.11.10 -> 4.11.11
<NixOS_GitHub>
nixpkgs/master b04858d Tim Steinbach: linux: 4.9.37 -> 4.9.38...
NixOS_GitHub has left #nixos []
<Infinisil>
spinus: How can I find out what modules are loaded?
<spinus>
lsmod
k0001 has quit [(Ping timeout: 240 seconds)]
<Infinisil>
Ahh
<Infinisil>
Alright I'll restart to see what it looks like while it works
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] NeQuissimus pushed 4 new commits to release-17.03: https://git.io/vQdFm
<NixOS_GitHub>
nixpkgs/release-17.03 05e8bb0 Tim Steinbach: linux: 4.12.1 -> 4.12.2...
<NixOS_GitHub>
nixpkgs/release-17.03 170e378 Tim Steinbach: linux: 4.11.10 -> 4.11.11...
<NixOS_GitHub>
nixpkgs/release-17.03 da2479a Tim Steinbach: linux: 4.9.37 -> 4.9.38...
NixOS_GitHub has left #nixos []
wizwizwizwiz has quit [(Remote host closed the connection)]
Infinisil has quit [(Quit: restarting)]
Infinisil has joined #nixos
pie__ has quit [(Remote host closed the connection)]
slyfox_ has joined #nixos
slyfox has quit [(Ping timeout: 240 seconds)]
newhoggy has quit [(Remote host closed the connection)]
slack1256 has quit [(Remote host closed the connection)]
<Infinisil>
Damnit, lsmod has exactly the same output before and after replugging
<spinus>
any logs when unplugging?
<spinus>
(usually journalctl -f is my friend)
pie_ has joined #nixos
pie__ has joined #nixos
pie_ has quit [(Remote host closed the connection)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] NeQuissimus pushed 1 new commit to master: https://git.io/vQdFQ
<NixOS_GitHub>
nixpkgs/master df929d6 Tim Steinbach: linux-copperhead: 4.12.1.a -> 4.12.2.a
NixOS_GitHub has left #nixos []
simukis has quit [(Ping timeout: 260 seconds)]
hyphon81 has quit [(Remote host closed the connection)]
ryanartecona has quit [(Quit: ryanartecona)]
zeus_ has quit [(Remote host closed the connection)]
zeus_ has joined #nixos
<Infinisil>
spinus: Plenty, but I'm not sure what to make out of it
<Infinisil>
Maybe it has something to do with "hotplug", of which I have no clue
<spinus>
Infinisil: sometimes udev is sending /setting some stuff on the device when you insert
<spinus>
if you do lsusb and grab the usb ids, you could search for it in udev rules
<spinus>
in /etc/udev/
<Infinisil>
spinus: Something like 05ac:9227?
<spinus>
Infinisil: actually, did it work before?
<spinus>
yes
<Infinisil>
It worked no problem after the restart
<spinus>
ah
erictapen has quit [(Ping timeout: 248 seconds)]
zeus_ has quit [(Ping timeout: 246 seconds)]
<spinus>
nice
<Infinisil>
No udev rule found
<spinus>
any upgrade? maybe new driver or something?
<Infinisil>
Don't think so
<Infinisil>
I did find a bunch of lines like this in journalctl: