<{^_^}>
#60869 (by jtojnar, 8 hours ago, open): libredirect: do not redirect dlopen by default
o1lo01ol1o has quit [Ping timeout: 258 seconds]
Jetien has quit [Ping timeout: 255 seconds]
<worldofpeace>
jtojnar: We know so far that it has only cased issues for sublime3 right?
<jtojnar>
yeah, but there are some other proprietary apps which might suffer from the same issue
<pie_>
grepping stderr is a lot nicer ever since i learned about |&
Glider_IRC_ has joined #nixos
Rusty1 has joined #nixos
<worldofpeace>
I really do hope you get a response to https://forum.sublimetext.com/t/how-is-dlopen-called-by-sublime-text/43998 eventually and then we could get a proper fix. I think disabling it is perfectly reasonable, and if we can't find someone to look at it within a few days I think we should just merge it. I tried to figure out the issue and didn't get further than you did.
<pie_>
worldofpeace, how much of an answer does one need to "how"
<pie_>
is sublime text hard to disassemble? xD
<pie_>
strace? xD
<jtojnar>
strace is weird
<pie_>
#8alternatively, who wants to deal with that i guess)
<worldofpeace>
yeah pie_ these aren't fun issues :)
<{^_^}>
commercialhaskell/stackage#4466 (by phadej, 3 weeks ago, open): Metadata contains wrong version for LTS-13.16 for transformers and process packages
<blackriversoftwa>
angerman: Yeah I'm trying to use the latest lts (13.19)
qubasa_ has joined #nixos
Qubasa has quit [Ping timeout: 255 seconds]
<blackriversoftwa>
angerman: Ah that explains it. Pretty simple matter to just bump those in the stack.yaml I'm working from until that issue is resolved
<angerman>
blackriversoftwa: I think if you set the extra-deps in the stack.yaml file to match that should solve it. Yep
drakonis has quit [Ping timeout: 276 seconds]
drakonis has joined #nixos
ardumont has quit [Ping timeout: 246 seconds]
<blackriversoftwa>
angerman: now I'm being treated to the soothing sound of a CPU fan as this build runs. Thanks for your help :)
<angerman>
blackriversoftwa: any feedback on what could be done better is much welcome :-)
<samueldr>
off-topic: ⚠️ hey, peeps, might want to disable firefox sync for now, or not start firefox for now if it's not started, there seems to be something going on with firefox syncing
<samueldr>
(not a nixos issue)
ris has quit [Ping timeout: 260 seconds]
<aanderse>
samueldr: :( i just noticed
<aanderse>
my firefox just freaked out on me
<aanderse>
unfortunately this might be the last straw...
<samueldr>
:(
<aanderse>
i hate google chrome because it isn't firefox, but my crappy work laptop chugs with firefox
<aanderse>
i tried google chrome and my cpu usage is down to nothing compared to what it was
<aanderse>
browsing is so much faster
<aanderse>
but i *hate* chrome!
<samueldr>
I'm a bit worried, considering all the "neat-o" alternatives like qupzilla and such are surely less secure than using the big bois browsers, due to the small time devs not being able to commit the security effort
<samueldr>
(understandably)
<ivan>
why do you hate chrome? (Chromium?)
<aanderse>
ivan: simply because it isn't firefox
<aanderse>
ha ha ha
<jackdk_>
Probably not trusting the big G, or its influence on web "standards"
<aanderse>
jackdk_: yeah those are true too... but not the reason i hate chrome
<samueldr>
I personally am fine with chrome, I practice browser agnosticity by using both chrome and firefox on different computers I use daily (work/personal)
<samueldr>
but so want to be able to keep using multiple
<samueldr>
also → #nixos-chat for further non-nixos talk (got pulled in too!)
<aanderse>
what do you use for bookmark+tab sync?
<aanderse>
....
<jackdk_>
(you're not in nixos-chat) I don't. I have different piles of tabs on different machines.
eraserhd has quit [Quit: WeeChat 2.3]
jackdk_ has quit [Remote host closed the connection]
jackdk_ has joined #nixos
drakonis1 has joined #nixos
<worldofpeace>
where does share/devhelp/references get multi outputed into devdoc?
<worldofpeace>
jtojnar: yeah only 'share/devhelp/books` gets `_multioutDocs`. probably should add `references` to that?
<jtojnar>
sounds good
<pie_>
gchristensen, \o/ :3
<samueldr>
in case users here aren't in #nixos-chat, there is a temporary solution in setting xpinstall.signatures.required to false; extension data should be safe if you haven't removed them
<rrr>
Any idea what option should I add, to some *kernelModules, to be able to mount a raid1 logical volume?
<rrr>
I'm trying everything I see around, without understanding much. And since I don't see errors while nixos-rebuilding, I reboot every time to check, and nothing works so far
attero[m] has quit [Ping timeout: 264 seconds]
<rrr>
Where's a list of the possible kernel modules btw?
attero[m] has joined #nixos
<clever>
rrr: lvm?, mdadm?
<rrr>
lvm
zaphar_ps[m] has quit [Ping timeout: 264 seconds]
<clever>
rrr: lvm is always available in the initrd
zaphar_ps[m] has joined #nixos
Glider_IRC_ is now known as Glider_IRC
drakonis has quit [Ping timeout: 258 seconds]
<clever>
rrr: what is your configuration.nix and what is the error it has?
thc202 has quit [Ping timeout: 245 seconds]
<nh2>
jtojnar: what did you mean with "Dependency changes described in release notes do not affect us"? I didn't understand that
<rrr>
I made a mirrorred thing. Maybe I did it wrong. But it's not activated, and when I try to `lvchange --activate y /dev/stuff`, the message I get is that can't process, "raid1 target support missing from kernel?"
xenofungus[m] has quit [Ping timeout: 264 seconds]
<jtojnar>
pie_: hmm, that is what we are doing in the test but LD_PRELOAD works there
<rrr>
My configuration.nix, regarding kernelModules, is volatile. I'm just trying permutations of good looking strings, in kernelModules, availableKernelModules, initrd.kernelMo, initrd.availableKe...
<clever>
rrr: you may want boot.initrd.availableKernelModules = [ "dm-raid" ];
<pie_>
jtojnar, i have no idea whats going on
<rrr>
clever: tried that
<rrr>
I'll keep it in
<rrr>
but there's something else missing
<clever>
rrr: lvm will activate on its own, so you shouldnt need to lvchange -a y
<pie_>
jtojnar, you also patched openat in the same patch? (not sure)
<jtojnar>
nh2: I meant that all the dependencies that were mentioned in the NEWS as removed, we did not use before
<rrr>
clever: All the other lvm drives are activated. It's just this one, which is the only one mirrorred
<jtojnar>
pie__: I tried removing everything but dlopen from libredirect and the issue still persisted
<pie_>
ok
<rrr>
I'll try that out
<pie_>
jtojnar, can you think of any way i can help
<rrr>
clever: is there a place where you are browsing the kernel modules to try?
<clever>
rrr: the find command i gave above
<jtojnar>
pie_: I have no idea how to proceed any more
<rrr>
AHh, missed it
<jtojnar>
just looking for ideas
<rrr>
clever: thanks
<rrr>
rebooting
rrr has quit [Quit: WeeChat 2.4]
<pie_>
jtojnar, so theres sublime3, and theres this rewrite library which overrides some stuff via ld_preload, and the sublime crashes with preload set?
<pie_>
jtojnar, youve removed everything but the dlopen override?
<pie_>
and sublime doesnt output the "failed to load" string
<clever>
pie_: ive noticed before, that chromium will deadlock if libredirect is in place
<jtojnar>
pie_: it just quits since it cannot find the libraries
lsyoyom has quit [Read error: Connection reset by peer]
<pie_>
its not intel PIN i suppose :P (some sort of instrumentation library)
<jtojnar>
pie_: it prints it for some libraries sometimes :-P
rrr has joined #nixos
<rrr>
Still doesn't work
<pie_>
jtojnar, disclaimer: im not super familiar with this stuff
dsiypl4 has quit [Remote host closed the connection]
<rrr>
The find command is indeed finding both dm-raid.ko.xz and dm-mirror.ko.xz
dsiypl4 has joined #nixos
dbmikus_ has quit [Ping timeout: 268 seconds]
<clever>
rrr: run both `modinfo dm-raid` and `modinfo dm-mirror`, then read the descriptions
<pie_>
this is the code that references one of the screenshotted functions, not quite sure whats going on https://bpaste.net/show/83a49d5d29ba , dunno what the "possible overlap" means either
<pie_>
jtojnar, any idea what the first lib that fails is
<nh2>
even though I have a `darwin.apple_sdk.frameworks.VideoToolbox` dependency.
<nh2>
` fatal error: 'VideoToolbox/VideoToolbox.h' file not found`
dsiypl4 has quit [Ping timeout: 246 seconds]
<pie_>
jtojnar, do you know if we should be overriding dlsym as well? i imagine not since it takes a dlopen handle
<pie_>
and that seems abstract enough
<jtojnar>
pie_: that is my thinking as well
jackdk_ has quit [Ping timeout: 244 seconds]
<nh2>
does include path resolution work somehow differently on Darwin, so that `VideoToolbox/VideoToolbox.h` is supposed to work even though `VideoToolbox.h` isn't in a directory called `VideoToolbox`?
<jtojnar>
nh2: maybe it is another gstreamer bug
<pie_>
jtojnar, my completely wild guess is this is going to be some arcane low level problem but idk :D
o1lo01ol1o has joined #nixos
<rrr>
I'll try with mdadm instead. Never did software raid before
<nh2>
jtojnar: I thought so too, but many projects have `#include <VideoToolbox/VideoToolbox.h>` in the source code, including ffmpeg and so on
<jtojnar>
nh2: or it was there previously and MacOS update moved it
<jtojnar>
sorry, I do not really know much about darwin
<pie_>
jtojnar, can you maybe make a reproducing single nix file that i can repro locally with
<jtojnar>
pie_: I tried creating libredirect testcase for dlopen but there it works fine
<pie_>
jtojnar, i meant with sublime
<jtojnar>
pie_: just nix-build -A sublime3-dev on master will work
<pie_>
im no low level expert but i dont see anything obviously wrong yet :I figured i could try throwing gdb at it too
<pie_>
the only path i see that doesnt hit the error message is where it succeeds at opening the dll but maybe fails to load a symbol
Supersonic has quit [Ping timeout: 258 seconds]
<nh2>
jtojnar: yea me neither, I think closed source OS are silly. I just got into fixing the Darwin because it looked like a low hanging fruit and would be cool if more people could use my gstreamer low-latency screen sharing pipeline. And now I got myself into that :o
<pie_>
but i could have misread
<pie_>
jtojnar, bit confused here though, man page says dlopen is a void function DAT_00ae8690 = dlopen("libcairo.so",1);
<pie_>
maybe its a disassembler artefact but that return value looks like it gets used as a status flag so i dont get it
jtojnar has quit [Remote host closed the connection]
<pie_>
well my C isnt that great, its actually of void pointer type so i guess it returns a pointer
jtojnar has joined #nixos
yastero_ is now known as yastero
lsyoyom has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
Supersonic has joined #nixos
<clever>
void *dlopen(const char *filename, int flag);
growpotkin has quit [Remote host closed the connection]
<nh2>
jtojnar: there must be some special thing about it, because `<Cocoa/Cocoa.h>` also doesn't exist this way on the file system, but that one was found in the nix build after I added the Cocoa dep. Will investigate more
silver has quit [Read error: Connection reset by peer]
<pie_>
jtojnar, i get a bunch of unable to load messages
<pie_>
jtojnar, what does rewrite do
rprije has joined #nixos
<jtojnar>
rewrite replaces the path with its counterpart from NIX_REWRITES env var
<pie_>
jtojnar, is rewrite a native libc function or is it something we added
<pie_>
because it could collide with something else if its something we added?
kreisys has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<jtojnar>
in case of libcairo.so, it returns the original path
<jtojnar>
pie_: it is `static const char * rewrite(const char * path, char * buf)`
<pie_>
jtojnar, my first message is Unable to load cairo_clip from libcairo.so which matches the first thing in this list https://bpaste.net/show/83a49d5d29ba
<jtojnar>
so it should not collide with anything
<pie_>
you said its not even getting that far?
<jtojnar>
pie_: no, I am getting the same thing
<pie_>
oh ok
<pie_>
that means its at least trying
<pie_>
jtojnar, seems to match the list perfectly
<jtojnar>
pie_: one difference I noticed is that strace shows different set of directories
<pie_>
jtojnar, running the loading code *is* conditional on DAT_00ae8698
<pie_>
i havent decided if its just a "ive loaded this alreday" flag yet (because im slow :P)
<jtojnar>
that's what I would guess
<pie_>
because right after the check it goes and sets it to 1
<jtojnar>
from the snipper
<jtojnar>
s/snipper/snippet/
<pie_>
but if it gets it as 1 already then *shrug* wouldnt make much sense though
<clever>
pie_: it could be a generic load function, that is ran from many places
<clever>
and should only act once
<pie_>
clever, thats what i figured
<pie_>
the static analysis shows that area as the only code that it can see as writing to that variable so i guess all known things point to it being fine
<pie_>
jtojnar, so either all the dlopen results are 0
<pie_>
jtojnar, or all the dlsyms are 0
<pie_>
jtojnar, can you confirm?
<jtojnar>
I would expect the dlsyms to be fine if the library was opened correctly
<pie_>
you can just printf the calls in your code for testing or something right?
<jtojnar>
so would expect dlopen to fail as strace cannot finde the file
<teto>
is it possible to specify a hash to builtins.fetchGit ? had no internet yesterday and I tried to change ttl options (btw tarball-ttl is not mentioned in man nix.conf)/setup a hash to use the source in cash but couldn't :'(
<jtojnar>
pie_: an idea, is sublime's runpath not inherited by ld_preloaded libraries?
<jtojnar>
pie_: that would explain the different set of searched directories
<pie_>
jtojnar, you have to explain more or give me a bit to process
lovesegfault has quit [Ping timeout: 252 seconds]
<jtojnar>
previously, sublime called `dlopen(libcairo.so)` and dynamic linker used sublime's rpath to find the library
<jtojnar>
now we call dlopen in our ld_preloaded libredirect, which has a different rpath, not containing `${cairo}/lib`, thus unable to find cairo
lovesegfault has joined #nixos
<pie_>
the dynamic linker uses rpath for libraries?
<pie_>
oh huge derp, of course it does
<pie_>
jtojnar, thats a very reasonable sounding hypothesis
<jtojnar>
now, how do we inherit sublime's rpath?
<pie_>
my next question besides testing would be how does it know if its being called from libredirect or from sublime
<pie_>
not quite the right question but
<pie_>
jtojnar, i started typing rpath ld_preload and google started suggesting things so this sounds very suspiciously on the right track
<jtojnar>
I guess some ELF magic, but that is outside my expertise
<pie_>
or not lol, ld preload and rpath being mentioned in the same context isnt such a hard thing to do
<pie_>
jtojnar, in the interim, can you write a test to get the rpath?
lovesegfault has quit [Ping timeout: 250 seconds]
lurpahi has quit [Ping timeout: 264 seconds]
lurpahi has joined #nixos
<dtz>
lol I just posted about it the rpath/etc. thing. In short, LD_PRELOAD intercepting dlopen() is difficult to get right and really really is not something you wnt to be responsible for unless your hobby project is literally writing a new dynamic linker or so
<pie_>
maybe you could have the preloaded library check what executable its running on, have it patchelf itself with the new rpath, and load itself again :D
<pie_>
dtz, also how the heck do you know about this :p
ardumont has quit [Ping timeout: 246 seconds]
<dtz>
TBH I really am skeptical that most programs do any of this in an intentional manner, since there are a number of edge cases unspecified and in the end they almost certainly kludged it until it stopped crashing and shipped it :P. Not sure I blame them, it's such a complete mess, and not many oracles really beyond "what does this implementation do" re:reasoning about what any of it should do
<pie_>
jtojnar, i forgot what the problem to solve was
<dtz>
i.... research these things, for reasons I am not sure were sound and often regret. ;P.
<pie_>
dtz, please indulge my curiosity :P
<pie_>
im thinking maybe malware people do something we can use
<dtz>
however deep you think the rabbit hole is, it's probably much much worse
<pie_>
yeah i have 10 glibc tabs open
<jtojnar>
dtz: I guess we will only intercept absolute paths in dlopen, for our sanity
<pie_>
im not sure this was a good idea
<dtz>
starting resources are drepper's DSO document-- read that and see how things evolved, helps things make sense.
<gchristensen>
dtz: glad to have you here with us :)
<pie_>
dtz, though i mean im curious how *you* got into this :p
<pie_>
this looks like a good doc, also the name sounds familiar for some reason
<dtz>
one of the best books on such, "Linkers and Loaders" has a joke in the preface about how it's really a funny situation as the book's audience is essentially the same 4 people interested at the level of such. Err, that's not quite right sorry I'm gonna get some food/caffeine and maybe find my copy ;).
<jtojnar>
I think I have it somewhere, still
<dtz>
yeah the dsohowto.pdf . It is not... an expalnation of how things work. Generally that's not really a documented situation--although i think that's largely changed "lately" (last years) so I don't mean to spread FUD. It's just a big mess, aways been a big mess, and despite being the fundamental way we compose the objects with which we build software... its entirely an undefined mess of insanity ;)
<pie_>
dtz, oh i started reading that book :3
ardumont has joined #nixos
<pie_>
dtz, sounds horrible, i want to write a decompiler with a formal spec
o1lo01ol1o has quit [Remote host closed the connection]
<pie_>
autogenerating specs for random software is probably somewhere on the uncomputable bucket list
<clever>
pie_: i saw something a year ago, about somebody writting the spec for a URL, purely by data from fuzzing a URL parser, lol
<clever>
this crazy guy took a windows driver, mutated it into an ELF library, and then dlopen'd it into his program
<clever>
so he could debug it with gdb :P
<pie_>
clever, just run it in wine tho
<pie_>
:p
<pie_>
oh uh kernel driver
<dtz>
yeah!!!! that's awesome stuff
<clever>
yep
<dtz>
guys why. WHY. we're just doing it to buildthe bugger right? this is .. unnecessary, surely? :( . How does anyone else build it? gah
alexherbo2950 has joined #nixos
<pie_>
jtojnar, come back so i can throw ideas at you
<pie_>
D:
alexherbo295 has quit [Ping timeout: 246 seconds]
<dtz>
lol halfway to building ourselves a new dynamic binary rewriter just to fix this project's build wonkiness?
<dtz>
lol
<jtojnar>
pie_: here
mabel has quit [Ping timeout: 246 seconds]
<pie_>
dtz, well thats my problem right now, i cant remember if i was ever told what problem this is supposed to fix
<pie_>
or if i forgot
<pie_>
jtojnar, how is the dlopen override supposed to fix the problem (in theory)
<pie_>
i mean what was the original problem
<pie_>
yes of course ;D <dtz> lol halfway to building ourselves a new dynamic binary rewriter just to fix this project's build wonkiness?
digitalgrease_ is now known as digitalgrease
<pie_>
nixos has The Most Advanced Tooling TM
<jtojnar>
pie_: well, I added dlopen support to libredirect so I could build introspection data of NetworkManager (they wanted to use installed path of the library during build so I needed to redirect it)
<jtojnar>
but that caused issue when sublime text tried to dlopen libcairo.so (a relative path)
<pie_>
how does this help with the networkmanager thing
<clever>
jtojnar: what if you LD_PRELOAD a libcairo.so?
<clever>
if you try to dlopen something already in scope, it just returns the existing handle
<pie_>
or you could have flags for enabling/disabling dlopen overrides per project >_>
<jtojnar>
clever: yeah, or sublime could use LD_LIBRARY_PATH instead of rpath
<pie_>
jtojnar, the reason i thought about it was i was still wondering if we could have the dlopen run in the correct context
<maurer>
pie_: Not applicable to sublime, but I know Glade based applications will use dlopen+dlsym to introspect on themselves and find functions to bind to buttons, since what functions are bound to what functions is specified in an xml file
<maurer>
pie_: It's possible sublime is doing something similar (e.g. external data file, needs to look up symbols by string as a result)
<pie_>
maurer, o/
<jtojnar>
pie_: the loader is weird. For example, how does it share libraries between processes?
<pie_>
jtojnar, i dont think a loader would share libraries between processes? not sure what you mean
<pie_>
maurer, i guesss
<jtojnar>
pie_: IIRC, it was promoted as one of the benefits of dynamic linking that each library only needs to be loaded once per system
<jtojnar>
hmm, maybe mmap takes care of that
<pie_>
jtojnar, libraries get loaded to different parts of the same virtual address space, but the kernel can deduplicate that somehow with memory mapping or something
<clever>
pie_: the loader uses mmap to map the library into a process
<clever>
pie_: and if 2 procs try to mmap the same region of the same file, in readonly mode, the kernel will reuse the pages
<pie_>
clever, makes sense
mabel has joined #nixos
<clever>
pie_: i think the kernel will even treat the ELF file sorta like a swap file for those pages
<maurer>
jtojnar: You are partially right - on windows, special care is needed
<maurer>
jtojnar: because it relocates the shared objects
<clever>
pie_: so it can just drop the .text in /dev/null when low on ram, and read it back on when its next needed
<maurer>
On linux though, it's as clever says. You just map the same file chunk in, and the kernel deals with it
<clever>
linux also requires you to build with PIE, so it can load at a different address and not need relocations
<clever>
not 100% sure what windows does/needs
<maurer>
(I mostly know about the windows stuff because this at was at least previously a weakness in their ASLR)
<pie_>
jtojnar, the networkmanager thing was too vague for me to get it
<maurer>
(Since the same dll would be loaded at the same address in every process if possible)
<clever>
maurer: *doh*
<clever>
maurer: linux only suffers from a form of that via fork() for worker children
<clever>
maurer: all children inherit the layout from the parent
<maurer>
Yeah, that's kind of unavoidable
<maurer>
well, it might be avoidable, but it would be *expensive* to avoid
<pie_>
jtojnar, i mean, if stdenv has a flag to set or a package with a setuphook i guess thats fine
<pie_>
jtojnar, but something something hella obscure things noone ever knows about 'xD
o1lo01ol1o has quit [Remote host closed the connection]
<jtojnar>
pie_: when NetworkManager is built, it tries to read a library in $out/lib (due to our gobject-introspection patching to absolute paths), but it is not installed yet
<jtojnar>
so we need a way to make it available somehow
<jtojnar>
hmm, I guess we could add a symlink in preBuild, instead of relying on libredirect
<pie_>
oh, that makes some sense
<pie_>
jtojnar, well yeah, ld_preload guts-fiddling sounds sketchier than a symlink
<pie_>
(i think?)
wfranzini has quit [Remote host closed the connection]
wfranzini has joined #nixos
<pie_>
leave the ld_preload for redirecting network requests to disk or something ;D
xcthulhu has joined #nixos
<pie_>
or writing portable applications that dont barf over your system or $LIST_OF_FUN_THINGS
<lambda-11235>
day|flip: Thanks.
<day|flip>
lambda-11235: no problem
<matthewbauer>
<nh2 "matthewbauer: Do you remember wh"> nh2: not sure but might just be to avoid too many differences with linux
<nh2>
matthewbauer: I tried to enable it just now and get all kinds of odd objective-c compile errors
justan0theruser has joined #nixos
justanotheruser has quit [Ping timeout: 255 seconds]
<pie_>
jtojnar, well, dlopen handles are opaque right so it might even be possible (theoretically) that its some kind of relative value so it *wont* work with another memory blob thing
<clever>
pie_: i think the handle is a pointer to the ELF header in ram
<clever>
pie_: that will apply an override to the stdenv, to enable debug
<pie_>
actually i guess i could give aswanson's reticulate bug another look
<pie_>
clever, yeah *a* problem was that didnt seem to work with rPackages
<clever>
rrr: indirect roots could become an issue if your sharing it like that
<pie_>
clever, and the rpackages code looked kind of bad to me too :( couldnt figure out how to override anything, so i ended up overriding the stdenv i passed in with some stuff set in preBuild
<pie_>
it was horrible and didnt work (probably because i was probably setting the wrong thing)
<rrr>
I don't know what indirect roots are. I'll assume it isn't safe then to bind mount the nix store on fedora for example.
<clever>
pie_: all enableDebugging does, is change the stdenv with .override
<pie_>
clever, yeah i looked at the source
<pie_>
clever, ah rigt
<pie_>
clever, wait wtf...why was i doing what i did then
<pie_>
:D
<pie_>
rpackages broke my brain
<pie_>
clever, oh right
<pie_>
clever, the problem is you need to apply enabledebugging to the result and not to the stdenv you pass in, because if you do it to the latter it will just get overridden
<pie_>
but that doesnt sound right, because override changes the inputs *shrug*
<pie_>
clever, what i actually did is override the pkgs.stdenv that gets passed in
* pie_
checks his code
<clever>
if rpackages accepts a { pkgs }:, it wont care about .override { stdenv = ?; }
<pie_>
yeah, i agree
<pie_>
but the theory should still work
<pie_>
ok so i think i did have dontstrip set correctly, but the optmization level not, so ill go peek through my code and see what i can do with my newfound stuff
<clever>
also, peek at the .drv that is building the thing
<clever>
nix show-derivation and nix-instantiate
ayerhart has quit [Quit: ayerhart]
<pie_>
clever, another fun thing, r doesnt use environment variables for builds, it has --config-vars --config-opts and some kind of makevars file
<pie_>
guess how many of these the rpackages infra exposes
<pie_>
well it does expose --config-opts
<pie_>
im not even mad, mostly just lost
<clever>
> pkgs.path
<{^_^}>
/var/lib/nixbot/nixpkgs/master/repo
ayerhart has joined #nixos
ayerhart has quit [Read error: Connection reset by peer]
<jtojnar>
pie_: just realized the rpath hypothesis does not explain why it worked when I just called the dlopen directly without calling rewrite
ayerhart has joined #nixos
<dtz>
uhhh so the results of running (likely incomplete but definitely catching real problems) this quick grep is.... discouraging: `rg '/nix/store/[0-9a-zA-Z]{32}-[^/]*/[[:alnum:]_\./-]*' /etc/ -L -o --no-filename|while read f; do [[ -e $f ]] || echo "$f"; done|less` (lol don't judge me i was just curious re:referencing paths that don't exist)
<dtz>
err rg is ripgrep if that's not your jam. Likely could just use grep -r lol
<dtz>
there are also a bunch of broken links QQ
fusion809 has quit [Read error: Connection reset by peer]
justanotheruser has quit [Ping timeout: 250 seconds]
marusich has joined #nixos
drakonis1 has quit [Quit: WeeChat 2.4]
dbmikus_ has joined #nixos
ayerhart has quit [Ping timeout: 268 seconds]
mounty has joined #nixos
domogled has joined #nixos
<lambda-11235>
I'm having a couple of problems. First, when I log into a non-root user PATH does not include /run/current-system/sw/bin. Second, when I do include that path sudo gives an error?
jtojnar has quit [Remote host closed the connection]
<clever>
lambda-11235: because sudo is in /run/wrappers/bin
justanotheruser has joined #nixos
<clever>
lambda-11235: did you programs.fish.enable = true; ?
<lambda-11235>
clever: No, I'll try that.
justanotheruser has quit [Ping timeout: 250 seconds]
justanotheruser has joined #nixos
init_6 has joined #nixos
<lambda-11235>
Thanks, that worked.
justanotheruser has quit [Ping timeout: 245 seconds]
wfranzini has quit [Remote host closed the connection]
wfranzini has joined #nixos
palo1 has joined #nixos
<pie_>
so i have a question
<pie_>
how does TOR fare with the firefox implosion
jchia1 has joined #nixos
<lopsided98>
pie_: I heard that noscript got disabled in Tor browser
<pie_>
i hope this doesnt end up having have long term negative repercussions for the mozilla projects and instead leads to bette things
<pie_>
:<
<jchia1>
I just install NixOS 19.03 on a laptop that I'll use to learn NixOS and develop software. I'll be installing the GUI next. I haven't decided which DM to use. Is it easy/advisable to install Wayland instead of X? I'm looking for a relatively low-maintenance experience compared to X. What's the status of Wayland support?
<jchia1>
The NixOS manual references services.xserver.enable for enabling X, but I can't find something that looks similar for Wayland here: https://nixos.org/nixos/options.html#wayland
<jchia1>
Also, what does the option services.xserver.displayManager.gdm.wayland mean? I thought wayland is a replacement for X, so why is it under 'xserver'?
<day|flip>
GNOME is still the only main DE that kind of head of thing. sway is i3 on wayland. but most wayland option. i don't know if there are any.
<day|flip>
if you go with gnome. you better off using gdm
<day|flip>
kde wayland. i don't know if it usable state like GNOME yet
aterius has joined #nixos
<pie_>
given our model does that mean we basically need to push out a firefox update ASAP once the fix is out and basically end up with mass user migration to master? :P
<day|flip>
are you talking about the addons bug?
<day|flip>
just go to about:config and change this to false. xpinstall.signatures.required
<pie_>
gee that doesnt sound sketchy at all
<pie_>
but thats not the point
<pie_>
im not actually affected yet
<day|flip>
will im am
<day|flip>
im using that for a temp work around. till fix pops up
<pie_>
my question is if this needs to be fixed on a distro level or if theyll be able to push an update that we can use
<pie_>
because we have immutability right, so a user cant just use the normal firefox update mechanism
<pie_>
s/right/right?/
<srhb>
pie_: Right.
<pie_>
well ok, i guess on a theoretical level this wouldnt be any different than a security update
<srhb>
No different from any update at all.
<day|flip>
im not seeing a package update for firefox in nixpkgs issuse
<pie_>
there isnt even a fix out yet
<day|flip>
to me it seem more of a cloud fix
<pie_>
i think im about to crash, ive been up all night anyway
<pie_>
also ive been thinking nix needs an update nag screen like windows
dansho has quit [Ping timeout: 250 seconds]
mexisme has quit [Ping timeout: 246 seconds]
pwnst has joined #nixos
<pwnst>
:q
pwnst has quit [Client Quit]
endformationage has quit [Ping timeout: 250 seconds]
<maurer>
pie_: god no, that defeats half the point for the non-user-software purposes
<maurer>
"Hey, have you considered updating libfoo and libbar?" "Go eat a dick. I'm using nix specifically so I keep getting the same shit until *I* say it changes"
<tilpner>
jomik: You can always try to throw buildFHSUserEnv at your problem. Usually it works
orivej has joined #nixos
throwson has joined #nixos
Dema has joined #nixos
throwson is now known as throwson2
throwson2 has left #nixos [#nixos]
<jomik>
tilpner: Hmm, isn't that sort of cheaty though ? :P
acarrico has quit [Ping timeout: 250 seconds]
<Myrl-saki>
jomik: Maybe one can make a setup hook.
<Myrl-saki>
I actually think there's a patchelf setup hook.
<tilpner>
jomik: Yes, and it's an expensive-ish solution. For nixpkgs it should only be used when absolutely necessary, but if you just want this for yourself...
<Myrl-saki>
/etc/nixos/nixpkgs/pkgs/build-support/setup-hooks/auto-patchelf.sh:1:declare -a autoPatchelfLibs
<jomik>
How do I enable that tho? :O
<jomik>
Ah, I think that works only for the things I move to $out.
<jomik>
It is so weird, this program. Because the thing we want to execute is in `./` and it is a shell script, which sets up some variables and then runs a `./bin/do-not-run*` binary. So I don't want to actually expose the stuff in the `bin` folder
<Dema>
Hi! I have a question, yesterday I fixed a bug in pygame derivation, now it's been merged to nixpkgs/master, what should I write in configuration.nix to force nix to use this derivation instead of the one in unstable channel? Tried adding this https://pastebin.com/E2t76U7E but it then fails to build another unrelated package with "wrapPython not found in buildInputs = [ curaengine python27Packages.wrapPython ];"
mexisme has quit [Ping timeout: 258 seconds]
<tilpner>
> :p { python27Packages.pygame = 42; }
<{^_^}>
{ python27Packages = { pygame = 42; }; }
<tilpner>
Dema: You are clearing out the rest of python*Packages
linarcx has joined #nixos
day|flip has quit [Remote host closed the connection]
<Dema>
This makes sense.. How should I override it then?. I'm a bit lost..
aexoxea has joined #nixos
<srhb>
Dema: I believe there's an example in the nixpkgs manual.
<tilpner>
Naively you could do { python27Packages = pkgs.python27Packages // { inherit (nixpkgs-master.python27Packages) pygame; }; }
<srhb>
Dema: Section 9.14.3.2
_e has joined #nixos
<tilpner>
But there sometimes are package-set specific override/extension functions, and I don't know if pythonPackages has one
<srhb>
It does, each python attribute takes a packageOverrides argument in the usual self: super: { ... } form
mexisme has joined #nixos
<srhb>
(But the naive version should probably be fine for just that single override, unless needed for other python packages...)
<srhb>
Dema: More generally you might also want to read the overlays section.
<jomik>
How do I access autoPatchElf in my nix expression fixup phase?
<srhb>
So 12.1.1 and 9.14.3.8 might be good.
<srhb>
jomik: Are you sure you need to do it manually? If you avoid overriding the default phases, everything should just work
<Dema>
Thanks!
<srhb>
jomik: (So, try to use eg. pre/postFixup instead of overriding fixupPhase)
<jomik>
srhb: It only patches in $outputs >.o
<jomik>
I need to touch stuff in `./lib` and `./bin` that I don't move to $out.
<jomik>
At least I am not sure how I would go about moving stuff to `$out/bin` without having it show up in my path afterwards :P
<jomik>
This package is weird..
<srhb>
jomik: Just delete it afterwards from $out? :-P
mounty has quit [Remote host closed the connection]
<jomik>
But then it wont be patched? :D
<aszlig>
jomik: you can do that by using autoPatchelf <file>
<jomik>
No reason to patch it if I delete it ? :P
<srhb>
I guess I didn't understand the question then :-P
mounty has joined #nixos
<jomik>
it says it cannot find `autoPatchelf` aszlig
<aszlig>
jomik: do you have autoPatchelfHook in nativeBuildInputs?
<jomik>
I do now.. :D
<immae>
hello ! I have a growing number of "collision between .. and .." in my system configuration, do I have a way to specify the wanted priority? note: I cannot modify the packages themselves, they are from nixpkgs, and added by modules I use so I cannot "avoid" the additions
<srhb>
immae: You can still modify the packages via eg. overlays
<immae>
is that a good solution?
<jomik>
srhb: I am trying to package the `Firestorm` second life viewer. And it is quite weird.. It has a bin directory, but you are not supposed to execute anything from that. You need to run a shell script from the "root" directory... And there is a lot of things to patch.
<srhb>
Possibly. Avoiding collissions in the first place is probably best, where possible.
<srhb>
immae: But that's certainly a way to specify priorities.
<aszlig>
jomik: so i guess it's for binaries you need to execute during the build process, right?
<immae>
srhb: ok, it won’t trigger a manual build, will it?
<jomik>
aszlig: Nopi
<srhb>
immae: I don't think meta.priority will, no :)
<immae>
cool
<immae>
Thanks srhb :)
<jomik>
It is for the final package to be executed by the user
<jomik>
It is prebuilt..
<aszlig>
jomik: okay, then you shouldn't need to execute autoPatchelf manually
<jomik>
Well, maybe I need to move more stuff to $out then.. :P
<srhb>
immae: So essentially, an overlay like... self: super: { hello = super.hello.overrideAttrs (old: { meta = old.meta or {} // { priority = ...; }; });
<aszlig>
jomik: it automatically recurses through all outputs and runs patchelf on shared objects and elf binaries
<jomik>
I just don't think I "can".. Because, it wants stuff in ./bin, which does not belong there really.
<aszlig>
jomik: ah, so the whole package expects a certain directory structure, right?
<srhb>
immae: And you could make a neat function for that if you need to do it a lot.
<aszlig>
jomik: so you probably need to either do more wrapping or use something like libredirect to bend some paths
<jomik>
Yeah. Arch even does something like making a wrapper script that cds to the package directory before executing the shell script that runs the binary...
<immae>
srhb: Yes, I’ll do something like that, thanks :)
<immae>
I never played with priorities yet, so, it’s a good time to start
<jomik>
Guess maybe I should just build it in Nix..
<immae>
srhb++
<{^_^}>
srhb's karma got increased to 54
<jomik>
autoPatchelf can not find `libstdc++.so.6` ..
<jomik>
Nah, too much of a pain in my ass to fix up `Firestorm`. Thought it'd be easy enough to just snatch what I needed from the AUR package. But there is too much weirdo structure.
<jomik>
Time to put the baby to bed XD Have a nice day, and thanks for trying to help.
jomik has quit [Quit: WeeChat 2.4]
<aszlig>
jomik: hm... i think i did package some second life viewer a few years ago, but not sure which one, lemme check...
<aszlig>
ah, damn, too late
_e has quit [Quit: Lost terminal]
<aszlig>
and it was firestorm X-D
_e has joined #nixos
Jetien has quit [Ping timeout: 245 seconds]
dukedave has quit [Ping timeout: 252 seconds]
_e has quit [Client Quit]
dukedave has joined #nixos
jchia1 has left #nixos [#nixos]
orivej has quit [Ping timeout: 246 seconds]
freeman42x has joined #nixos
_e has joined #nixos
orivej has joined #nixos
ar1a has quit [Ping timeout: 268 seconds]
lambda-11235 has quit [Quit: Bye]
marusich has quit [Remote host closed the connection]
ardumont has quit [Ping timeout: 250 seconds]
ar1a has joined #nixos
<srhb>
,tell jomik Check the channel history, aszlig did package firestorm at one point. :-)
<{^_^}>
srhb: I'll pass that on to jomik
orivej has quit [Ping timeout: 244 seconds]
<aszlig>
srhb: ah, thanks, didn't know it has memoserv-functionality
<srhb>
It's very nice indeed :-)
<aszlig>
,tell jomik also, scrap that, i tried to do the source build but gave up on it at some point as well
<{^_^}>
aszlig: I'll pass that on to jomik
ardumont has joined #nixos
dbmikus_ has quit [Ping timeout: 244 seconds]
<cocreature>
I have some derivation where calling overrideAttrs seems to throw certain attributes away. Specifically, if I replace "drv" by "drv.overrideAttrs(oldAttrs: {})" I get an error about a "tlType" attribute missing later on (it’s a texlive package). Is that expected? Can I avoid this somehow?
civodul has joined #nixos
bobsterman[m] has joined #nixos
<immae>
srhb: while reading some modules to see why some packages are needed I saw lib.lowPrio and lib.hiPrio (in lib/meta.nix) which can help (not completely though, since it sets priority to ±10 so it doesn’t avoid clashes completely)
<srhb>
immae: Right, you can always set it yourself instead of bumping it like that.
<srhb>
Technically using setPrio is probably more right than manually wrangling the meta attr though, for portability.
knupfer has joined #nixos
<immae>
hmm I don’t see a setPrio?
<immae>
only hi and low
<immae>
ah right, in 19.03 it is
mounty has quit [Ping timeout: 245 seconds]
domogled has joined #nixos
Sabbat[m] has joined #nixos
Dema has quit [Ping timeout: 256 seconds]
Jetien has joined #nixos
Guest55252 is now known as jbetz
domogled has quit [Ping timeout: 245 seconds]
<enteee>
rxvt_unicide-plugins: how do I install such a plugin that rxvt_unicode finds them? Just installing both in the same environment does not seem to work.
domogled has joined #nixos
Ariakenom has joined #nixos
vidbina has joined #nixos
sigmundv_ has joined #nixos
<genesis>
how to install a header like optparse.h ?
adamantium has joined #nixos
domogled has quit [Ping timeout: 255 seconds]
hedning_ has quit [Remote host closed the connection]
adamantium has quit [Client Quit]
talqu_ has joined #nixos
adamantium has joined #nixos
talqu__ has joined #nixos
<genesis>
let's try $out/include
mabel has quit [Ping timeout: 246 seconds]
AluisioASG has quit [Ping timeout: 259 seconds]
mabel has joined #nixos
adamantium has quit [Quit: Konversation terminated!]
adamantium has joined #nixos
sb0 has joined #nixos
Anton-Latukha has joined #nixos
Dema has joined #nixos
domogled has joined #nixos
englishm has quit []
mexisme has quit [Quit: WeeChat 2.4]
englishm has joined #nixos
domogled has quit [Ping timeout: 246 seconds]
vidbina has quit [Ping timeout: 244 seconds]
thc202 has joined #nixos
Wizek has quit []
Wizek has joined #nixos
acarrico has joined #nixos
<genesis>
i still have issue with fetchSubmodules :/
__monty__ has joined #nixos
<enteee>
ok I found the solution to my problem: I had to install pkgs.rxvt_unicode-with-plugins instead of just pkgs.rxvt_unicode. I got confused because defalt plugins did work..
Jetien has quit [Ping timeout: 246 seconds]
ardumont has quit [Ping timeout: 250 seconds]
<enteee>
genesis: what are you trying to achive?
<genesis>
seems if i change rev to a non tag it's fetching
<genesis>
i try to update z88dk
fgaz has quit [Ping timeout: 264 seconds]
fgaz has joined #nixos
fgaz is now known as Guest14210
atopuzov[m] has quit [Ping timeout: 264 seconds]
atopuzov[m] has joined #nixos
buffet_ is now known as buffet
ardumont has joined #nixos
vidbina has joined #nixos
GiGa has joined #nixos
ris has joined #nixos
<GiGa>
Hello, I'm trying to work on issue #60433 following a tip from @tadfisher
calbrecht[mb] has quit [Remote host closed the connection]
calbrecht[mb] has joined #nixos
<yuken>
Hmm. Godot seems to be behind on NixOS, guess that'll be my first "custom" package.
sb0 has quit [Quit: Leaving]
ng0 has joined #nixos
colemickens_ has quit [Quit: Connection closed for inactivity]
<tilpner>
> godot.name
<{^_^}>
"godot-3.1.1"
<tilpner>
Which version do you want?
kisik21 has joined #nixos
alexherbo29509 has joined #nixos
alexherbo2950 has quit [Ping timeout: 246 seconds]
BinkyTheClown has quit [Ping timeout: 264 seconds]
<immae>
srhb: after some clean up on my side, it turns out that most of the collisions are actually collisions coming only from `nixos/modules/config/system-path.nix`, so I guess I’m not the only one "touched" by this issue? Should I instead fix it there rather than only locally?
<immae>
(for instance coreutils + procps + utillinux alone make more than half of the collisions)
* tilpner
wouldn't mind a way to thin out requiredPackages
<immae>
well they seem quite important and minimalist to me (what would you remove tilpner?), but having reasonnable priorities seems like the job of that package
<immae>
s/package/module/
<tilpner>
immae: nano, perhaps strace, ssh, rsync
<immae>
ok
<immae>
makes sense (except for nano)
<gchristensen>
gotta have an editor :)
<immae>
(an editor is important yes)
<tilpner>
As a default, sure!
BinkyTheClown has joined #nixos
<tilpner>
But what's wrong with allowing a user to decide against the default?
joapet99 has joined #nixos
<tilpner>
Right now you can't really remove anything from that list
<joapet99>
Hewwo
<tilpner>
Short of disabledModules
<immae>
right
<immae>
(my initial point was setting correct priorities to avoid warning messages though, not adding or removing tools from there :D )
<gchristensen>
tilpner: how about a "noDefaultPackages" option or something
<kisik21>
SSH installed by default is the coolest thing in NixOS ever.
<immae>
kisik21: if you use a local machine where you don’t need ssh, your argument doesn’t work
<tilpner>
gchristensen: I want most of them though, and I don't want to copy that list to my own config and risk it getting out of sync
<gchristensen>
yeah but _I_ want none of them
<kisik21>
Well, I use SSH everyday :3 so for me it's the coolest. Coolness is subjective after all.
<immae>
kisik21: I do too, but it’s not the case of everyone.
<kisik21>
I specifically formulated it as a subjective. Sorry if it wasn't clear enough :3
<immae>
How about something like `withFoo` in configuration?
<tilpner>
gchristensen: Would everything still work without these? Do you know if others parts of NixOS implictly rely on them?
<immae>
kisik21: ok I didn’t take it as is, sorry for the misunderstanding :)
<enteee>
genesis: i think this is intended, because tags can move.
mounty has joined #nixos
<gchristensen>
tilpner: dunno but I'm willing to find out :)
<immae>
There are a lot of commands in activationScript that assume some of them present at least
domogled has joined #nixos
<immae>
(like "basic activationScript", not even custom ones)
<joapet99>
So I'm in a bit of a nix package mangement problem. I have problems launching Star Citizen in Wine with DXVK. DXVK though says that LLVM version 7.0.1 may cause problems so I thought I would try to change the LLVM version to either 7.0.0 or 8.0.0. Turns out it's not that easy on nixos and nixpkgs
<gchristensen>
yeah but anything but considering that a bug is wrong
<tilpner>
immae: activate sets a PATH at the top
<immae>
Right, but based on that list, no?
<immae>
Ah no distinct list, sorry
jasongrossman has quit [Read error: Connection reset by peer]
<tilpner>
immae: Your actual question was lost a little. Perhaps it would be a good idea to create an issue where you can list all collisions and which package you think should be prioritised?
wfranzini has quit [Remote host closed the connection]
<immae>
Well, I actually have no opinion on which should be prioritized
<immae>
but yes I can create an issue
wfranzini has joined #nixos
<immae>
the sensible choice would be "the one that nix choses currently", for least surprise
fusion809_ has quit [Remote host closed the connection]
fusion809_ has joined #nixos
domogled has quit [Ping timeout: 246 seconds]
knupfer has quit [Ping timeout: 264 seconds]
mounty has quit [Ping timeout: 258 seconds]
<talqu_>
hi, im trying to deploy trivial example from nixops manual (to a nixos machine). However i get these errors: - The ‘fileSystems’ option does not specify your root file system. - You must set the option ‘boot.loader.grub.devices’ or 'boot.loader.grub.mirroredBoots' to make the system bootable.
<talqu_>
infinisil: with this https://pastebin.com/Du44YaNT, I now get "You must set the option ‘boot.loader.grub.devices’ or 'boot.loader.grub.mirroredBoots' to make the system bootable."
<infinisil>
talqu_: Ah, I guess that stuff usually isn't in hardware-configuration.nix
<infinisil>
talqu_: Managing a NixOS system with Nixops requires the exact same configuration.nix as you would've used without nixops
<talqu_>
weird that they don't mention any of this in nixops manual
acarrico has quit [Ping timeout: 250 seconds]
<talqu_>
infinisil: i've added imports = [ ./configuration.nix ]; to trivial-env.nix where configuration.nix is the one of my remote machine and it worked :)
<talqu_>
so now i can move the systemd service logic to trivial.nix and finally have a deployment process, where the build happens on my local machine and then the resulting closures/binaries get copied to remote machine. Am i correct?
<infinisil>
talqu_: Yup :) I didn't suggest nixops originally because it's a bit overkill
<talqu_>
i added imports = [ ./configuration.nix ]; where configuration.nix is from my local machine and it also work.
freeman42x has quit [Read error: Connection reset by peer]
<talqu_>
can I add an openssh.authorizedKeys.keys for root via configuration.nix? Is it users.users.root = {...}?
pmiddend has joined #nixos
<pmiddend>
In order to use the Steam overlay with my game, I have to LD_PRELOAD an .so file. When I use the one from my $HOME directory, the app doesn't start because it doesn't find libGL.
<pmiddend>
Does anyone have experience with this?
<ben>
talqu_: id expect it to work
Glider_IRC_ has quit [Remote host closed the connection]
kl3 has joined #nixos
Glider_IRC_ has joined #nixos
<kl3>
how can i report bugs without a github account?
<ben>
kl3: I'd suggest this channel and hoping someone else posts them to github for you
<kl3>
nix-channel --update throws ENOENT on some /nix/store/... path. i have been using the unstable channel for the last few generations, rebooted into the latest, did `nix-collect-garbage -d' at some point, now this
<kl3>
should i open a topic on discourse? don't want to pollute this channel
o1lo01ol1o has quit [Remote host closed the connection]
srid has quit [Ping timeout: 250 seconds]
jonge has quit [Quit: quit]
jonge has joined #nixos
o1lo01ol1o has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
<pie_>
maurer, i didnt say the nag screen should force you to update :p
Xal has quit [Ping timeout: 246 seconds]
Xal has joined #nixos
<pie_>
so how do i get the firefox hotfix
* pie_
turns on "You can check if you have studies enabled by going to Firefox Preferences -> Privacy & Security -> Allow Firefox to install and run studies."
* pie_
waits for magic to happen
orivej has joined #nixos
<pie_>
yeah so nothing is happening
iqubic has joined #nixos
<iqubic>
So Mozilla screwed up and now none of my Firefox addons work.
<{^_^}>
#60916 (by deliciouslytyped, 13 seconds ago, open): Firefox 1548973 Hotfix (addons disabled due to expired intermediate certificate)
ambro718 has joined #nixos
init_6 has quit []
Kiren has joined #nixos
ilikeheaps has joined #nixos
kl3 has joined #nixos
kl3 has joined #nixos
kl3 has quit [Changing host]
hedning_ has joined #nixos
dsiypl4 has quit [Ping timeout: 252 seconds]
rpg has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
rpg has joined #nixos
rpg has quit [Remote host closed the connection]
lsyoyom has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
pmiddend has joined #nixos
waleee has joined #nixos
<pmiddend>
A library I'm using needs libstdc++.so.6. I find this in my /nix/store under /nix/store/...-gcc-7.4.0-lib. What's the corresponding package, though?
<gchristensen>
,locate lib libstdc++.so
<gchristensen>
,locate libstdc++.so
<{^_^}>
Found in packages: libstdcxx5, gcc-unwrapped.lib
<Kiren>
gchristensen: No, unfortunately, that's what I'm checking.
<gchristensen>
what does `systemctl status firewall` say?
petercommand has joined #nixos
dsiypl4_ has joined #nixos
<petercommand>
Hi, I am trying to install nixos via usb in my notebook (acer swift5), but I got a static cursor after booting from the graphical live CD iso
<petercommand>
I can't even get to the boot menu
<petercommand>
is there any way to debug this?
xok has joined #nixos
<xok>
hello all...
dsiypl4 has quit [Ping timeout: 246 seconds]
<xok>
I am trying to run a nixos deploy command against a remote machine...
<xok>
buut I'm getting "error out of memory" messages...
<xok>
I tried to add swap space but it doesn't seem to help...
<xok>
can anyone help with this issue?..
<Kiren>
gchristensen: Not familiar with this log format, but it says `Active: active (exited)`. And then there's a log containing starting events from when the machine was booted.
<Kiren>
gchristensen: Running `systemctl reload firewall` has not been making any differance either...
<gchristensen>
when you ran nixos-rebuild test, did it show it rebuilding a firewall file?
<xok>
does anyone know how to force nixops to go to swap if there's not enough memory?..
dansho has joined #nixos
slyfox has joined #nixos
Kim has quit [Quit: WeeChat 2.2]
erictapen has joined #nixos
Kiren_ has joined #nixos
srid has joined #nixos
Kiren has quit [Ping timeout: 256 seconds]
<xok>
never mind, I had to create swap on a target machine...
Kim has joined #nixos
<kl3>
i remember seeing nix crash due insufficient memory (2G), so that might have been the cause for old files lingering around
simendsjo has joined #nixos
drakonis has joined #nixos
slyfox has quit [Quit: no need to cry]
<xok>
kl3: I've cleaned the system but it still ran out of memory...
<xok>
creating the swap space helped ...
<Jonathan_>
In nixpkgs manual it talks about namespaces and that Hackage pkgs are only in the haskel.packages namespace... How do I use namespaces with declarative package management?
drakonis_ has quit [Ping timeout: 246 seconds]
<Kiren_>
gchristensen: Sorry for dissappearing. Had some trouble getting the build logs since I use a script to build.
<Kiren_>
Apparently I had changed `nixos-rebuild test` for `nixos-rebuild build` and running `result/activate` But apparently that was not the same thing.
<Kiren_>
rebuild test did reload the service properly. Thanks a lot for the help!
<gchristensen>
great!
<Kiren_>
Jonathan_: What are you trying to achive? Are you modifying your `configuration.nix` or creating a `shell.nix`?
<Jonathan_>
configuration.nix
shibboleth has joined #nixos
knupfer has quit [Remote host closed the connection]
<Jonathan_>
In nix-env packages can be install with `nix-env -iA haskell.packages.<pkg>`. It is touted as better/faster and it says some packages must be accessed with its attribute name.
<Kiren_>
Do you want a ghc with specific dependencies?
knupfer has joined #nixos
<gchristensen>
anyone familiar with how hydra's evaluator creates GC roots for every drv?
<tilpner>
Jonathan_: And pkgs.haskell.packages.<pkg> doesn't work?
<Kiren_>
Jonathan_: So, I'm pretty new to nix myself. As I understand it, you will have to install ghc bundled with the specific dependencies.
<Kiren_>
I'd reccomend reading Nix Pills on the _override pattern_, if you haven't already, to get a good sence of what that means.
<LnL>
gchristensen: hmm interesting, it seems to create empty files with the drv basename
<Jonathan_>
I am still trying to figure stuff out (:. I am trying to get the hackage package `http://hackage.haskell.org/package/implicit` (a cli app). I am also trying to set up my Haskell environment for development. But, also, simply trying to gain some understanding of how nix works.
<aszlig>
gchristensen: what's the issue?
<gchristensen>
r13y.com fails because its drv's get GC'd :)
<Jonathan_>
No pkg.haskell.packages.<pkg> does not work.
<aszlig>
gchristensen: do they get GCed *while* evaluating?
<gchristensen>
while building
<Jonathan_>
Well, I'll play with it some more, thanks for the help Kiren_.
<aszlig>
gchristensen: and you have set --gc-roots-dir?
<gchristensen>
yeah so it creates a GC root for the top level drv, but all the others get GC'd
<gchristensen>
I suppose I should just toggle that option ...
ardumont has quit [Ping timeout: 252 seconds]
<aszlig>
gchristensen: ah, okay, well, you probably need to set gc-keep-derivations then
<Kiren_>
Jonathan_: No worries! Since it's a bit of a hastle to change your environment each time you work on a different project I'd allso reccomend looking into `nix-shell`
<gchristensen>
yeah
<aszlig>
ah, that got renamed, it's called just keep-derivations
Kim has quit [Quit: WeeChat 2.4]
talqu_ has quit [Remote host closed the connection]
talqu__ has quit [Remote host closed the connection]
<Kiren_>
Jonathan_: Basically you create a `shell.nix` file. And then run `nix-shell` and it will provide you with this specific environment for that project.
Kim has joined #nixos
<Jonathan_>
Yeah, I remember bumping into that before, I'll dig it back up, thanks.
<Jonathan_>
What code editor do you use Kiren_? My favorite is TextAdept but it is not working on nixos right now.
shabius has joined #nixos
waleee has quit [Quit: WeeChat 2.4]
dsiypl4_ has joined #nixos
mabel has joined #nixos
Kim has quit [Quit: WeeChat 2.4]
ardumont has joined #nixos
Kim has joined #nixos
<Kiren_>
Jonathan_: What's not working about it?
<Kiren_>
I switched to only using vim recently to get more proficient with it. It's a bit of a hassle to get to configure on nix to start, but now I'm pretty happy with it.
<Jonathan_>
It `segmentation faults` when I try running it.
drakonis has quit [Ping timeout: 252 seconds]
lambda-11235 has joined #nixos
drakonis has joined #nixos
vidbina has quit [Ping timeout: 252 seconds]
ardumont has quit [Ping timeout: 245 seconds]
Kiren_ has quit [Ping timeout: 256 seconds]
<lunik1>
Has anybody set up ccache on NixOs? I've enabled it on configuration.nix and added "ffmpeg-full" to programs.ccache.packageNames but then ffmpeg compilation fails with "gcc is unable to create an executable file."
dsiypl4_ has quit [Ping timeout: 252 seconds]
ferriswheel has quit [Quit: Leaving]
o1lo01ol1o has quit [Remote host closed the connection]
<nh2>
jtojnar: what shall I do about the jack -> qt dependency, make jack optional?
<jtojnar>
nh2: that is probably the easiest way
<jtojnar>
in the fuure we should probably split ffado
o1lo01ol1o has joined #nixos
shibboleth has quit [Quit: shibboleth]
<steveeJ>
samueldr: I'm currently updating https://github.com/NixOS/nixpkgs/pull/55531 and I retested the printer (granted I'm not using it very often). I noticed that even when printing the hardcoded paper size (A4) it cuts something off at top of the page. any clues why that could be?
<steveeJ>
samueldr: ha I see. I figured you might know a trick since you have that paper size fix in another derivation
<samueldr>
I literally stumbled upon a solution when angrily wondering why my prints were cut at weirdly specific locations :S
Rusty1 has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
<nh2>
jtojnar: done
<steveeJ>
samueldr: gotcha, well. I'll try to find what's causing this here. I think we can still continue on the PR though. thanks for the review! I addressed your comments
dsiypl4_ has quit [Ping timeout: 246 seconds]
linarcx has joined #nixos
zeta has joined #nixos
<zeta>
when i try to do a nixos-rebuild switch, it says: `haskell-overridez` is broken, how do i fix this ?
<Philonous>
What's the difference between ststem.userActivationScripts and system.activationScripts ?
<Philonous>
I.e. when would I use which one?
<zeta>
is it just broken in nixos version 19.03 or is it the package itself ?
<clever>
ajs124: 7.4.0/crtend.o: relocation R_X86_64_32 against `.ctors' can not be used when making a shared object; recompile with -fPIC
<clever>
musl-ld: final link failed: nonrepresentable section on output
<ajs124>
Yeah, I saw those, but couldn't understand them.
<ajs124>
What is a nonrepresentable section?
<clever>
not really sure about them either
<steveeJ>
samueldr: have you tested the brother drivers on non-nixos to see if they're broken there too?
<ajs124>
How does position independent code factor into this?
<ajs124>
(which I assume is what PIC stands for)
<samueldr>
steveeJ: I don't have a non-nixos kicking around, otherwise I would have
<clever>
ajs124: not sure
<samueldr>
steveeJ: I also wonder how it works outside nixos
<steveeJ>
samueldr: is it a fact that it does? I thought we're just changing script paths
<samueldr>
steveeJ: not sure I follow?
<samueldr>
steveeJ: but maybe their install allows some impure modification in the FS
<steveeJ>
samueldr: are you assuming that the page size works correct on non-nixos or do you know for a fact?
<samueldr>
I assume, no evidence
o1lo01ol1o has joined #nixos
drakonis has joined #nixos
<steveeJ>
it's hard to find enthusiasm for debugging cups and proprietary drivers /o\
<samueldr>
I do agree
<samueldr>
I think, haven't completely validated, that when cups detects the browser through avahi/bonjour, in my case, it seems fine, I printed earlier this week without having my driver installed
stranger___ has quit [Quit: Connection closed for inactivity]
<samueldr>
maybe I accidentally wiped the right part of the print
<samueldr>
I just checked and it's fine...
drakonis_ has quit [Ping timeout: 252 seconds]
<samueldr>
I'll chalk it up to luck with a networked printer o.O
o1lo01ol1o has quit [Ping timeout: 264 seconds]
<steveeJ>
any chance cups itself is to blame on nixos?
<samueldr>
I'm willing to bet your knowledge of the situation is about equal to mine, steveeJ
<samueldr>
if not greater
cybrian has joined #nixos
dsiypl4_ has joined #nixos
drakonis_ has joined #nixos
npmccallum has joined #nixos
drakonis has quit [Ping timeout: 264 seconds]
shibboleth has joined #nixos
drakonis has joined #nixos
<ambro718>
Is there someone who is running NixOS in virtualbox here and has kernel 4.19? Can you check if "modprobe vboxsf" results in errors?
<ambro718>
I am trying to update VirtualBox to 6.0.6 and see strange problems loading vboxsf module.
xok has quit [Quit: Leaving.]
<ambro718>
It might be an issue not at all related to the upgrade but to the kernel gaining its own vboxguest and vboxvideo modules (and no vboxsf) which are loaded instead of the ones from additions, then vboxsf cannot work with those.
drakonis_ has quit [Ping timeout: 248 seconds]
jtojnar has quit [Quit: jtojnar]
janneke_ has joined #nixos
janneke has quit [Ping timeout: 246 seconds]
Myhlamaeus has quit [Ping timeout: 250 seconds]
sindrip has joined #nixos
sindrip has quit [Client Quit]
eraserhd has joined #nixos
o1lo01ol1o has joined #nixos
pointfourone has joined #nixos
pointfourone has quit [Client Quit]
o1lo01ol1o has quit [Ping timeout: 252 seconds]
<lambda-11235>
I'm trying to create a package, but I'm confused as to what prefix and DESTDIR should be set to. I've gathered that prefix should be $out, but what should $DESTDIR be?
<lambda-11235>
Does it just install to root, so DESTDIR should be left blank?
teh[m] has quit [Ping timeout: 264 seconds]
teh[m] has joined #nixos
mabel has quit [Ping timeout: 255 seconds]
<ambro718>
lambda-11235: for autoconf packages, yes I think prefix is usually set to $out and DESTDIR is not defined
<ambro718>
lambda-11235: usually you just let the default build steps call configure and they take care of many common options
<lambda-11235>
ambro718: Thankss. The thing I'm packaging only has a makefile that you pass prefix and DESTDIR to directly.
<ambro718>
by default, packages defined by mkDerivation make an effort to automatically do the right build steps, you don't call configure and make literally
<ambro718>
okay then
<lambda-11235>
Also, where can I find a list of the licenses in stdenv.lib.licenses?
FRidh has quit [Quit: Konversation terminated!]
<ambro718>
lambda-11235: lib/licenses.nix
<lambda-11235>
Ah, sorry. I thought was in the nixpkgs root when I was actually in ./pkgs. No wonder I had trouble finding it.
<pie_>
clever, (or anyone else that might be able to help) so i have this single-file reproducer for a bug, i cleaned it up a bit, still cant get debug symbols to work https://bpaste.net/show/07a1823756f1
<pie_>
gdb always says No symbol table info available. if i try to do something like info args
<pie_>
objdump -S on /nix/store/gv39cd8g28p39sig8bbp85bi5z5jxlqa-r-reticulate-1.10/library/reticulate/libs/reticulate.so doesnt show source code lines, which i heard should be a thing
evil-olive has quit [Quit: leaving]
gagbo has quit [Ping timeout: 250 seconds]
xkapastel has joined #nixos
mabel has joined #nixos
ambro718 has quit [Quit: Konversation terminated!]
bgamari has quit [Remote host closed the connection]
bgamari has joined #nixos
ambro718 has joined #nixos
m1cr0man has joined #nixos
<ambro718>
Shouldn't /run/opengl-driver not exist and instead only /run/opengl-driver-64 and /run/opengl-driver-32? Currently a 32-bit app would need to look both in /run/opengl-driver (to support 32-bit system) and /run/opengl-driver-32 (to support 64-bit system), which is inefficient as on 32-bit system it would always find unusable 64-bit libraries in the former and fall back to the latter.
talqu_ has joined #nixos
talqu_ has quit [Remote host closed the connection]
vidbina has quit [Ping timeout: 246 seconds]
<ambro718>
How do I prevent removal of a RPATH that I added intentionally?
<ambro718>
I added a RPATH with patchelf but it gets removed by "shrinking RPATHs of ELF executables and libraries"
<{^_^}>
#22760 (by nihil-2019, 2 years ago, open): VirtualBox cannot access OpenGL with Nvidia Drivers as Normal User
<ambro718>
but the automatic rpath shrinking gets in the way!
<lambda-11235>
I have a list of sources with name, sha pairs. Like so https://www.paste.org/98492. Nix is complaining about not being able to coerce a set to a string, why would that be?
<infinisil>
ambro718: dontStrip = true
dsiypl4_ has quit [Ping timeout: 250 seconds]
<infinisil>
Or not, something like that
zeug has quit [Quit: WeeChat 2.4]
<ambro718>
ah it's dontPatchELF
<infinisil>
Ah
<ambro718>
not optimal though since other RPATHs could be stripped
talqu has joined #nixos
<ambro718>
and that flag turns off this feature for ALL programs
<ambro718>
would it be possible to run my patchelf command after this stripping?
knupfer has quit [Ping timeout: 252 seconds]
<tilpner>
lambda-11235: That's just a set, not a derivation, and certainly not a source
nD5Xjz has joined #nixos
<tilpner>
lambda-11235: Perhaps you want to call a fetcher? They will need more arguments though
<lambda-11235>
tilpner: Later in the file I map a fetcher over each argument. I'll modify the post to give a clearer image of this.
<infinisil>
ambro718: postFixup = '' ...
<ambro718>
indeed, thanks
Rusty1 has quit [Quit: Konversation terminated!]
civodul has quit [Quit: ERC (IRC client for Emacs 26.2)]
<lambda-11235>
I've updated it. Should I just post the entire derivation (it's a bit long_?
jackdk_ has joined #nixos
talqu- has joined #nixos
<tilpner>
lambda-11235: No, what you pasted was correct
<tilpner>
As-in, it instantiates here
<tilpner>
So your error must be caused by something else
<tilpner>
lambda-11235: Currently, it's trying to provide submodules as an environment variable to the build process, but failing miserably because it doesn't know how to convert a list of attr-sets to a string
<roconnor>
I do not have nix-store locations in my path. But maybe there nixos managed user confurations whose profiles are stored in the nix-store?
<tilpner>
s/statement/expression/
dsiypl4_ has joined #nixos
<tilpner>
aswanson: One good reason would be that you're in a nix-shell
<aswanson>
tilpner: I was, but not any longer. The locations are specific packages like python and ranger
<aswanson>
its weird
adamantium has joined #nixos
<aswanson>
they are all things I have set to run when I log in, or dependencies thereof, I think
camsbury- has joined #nixos
gagbo has joined #nixos
ambro718 has quit [Quit: Konversation terminated!]
justanotheruser has quit [Ping timeout: 245 seconds]
domogled has joined #nixos
mabel has quit [Ping timeout: 248 seconds]
<illegalprime>
hey all, I can't seem to cross compile using a different gcc version. I'm trying this: nix-build -E 'with (import <nixpkgs> {}).pkgsCross.armv7l-hf-multiplatform; hello.override { stdenv = overrideCC stdenv gcc48; }
linarcx has quit [Quit: WeeChat 2.4]
mabel has joined #nixos
ardumont has quit [Ping timeout: 252 seconds]
wfranzini has quit [Remote host closed the connection]
wfranzini has joined #nixos
<lopsided98>
I installed static Nix on some machines that I don't have root access to, so I am using the --store option to put the store in a different directory
<lopsided98>
I wrote a wrapper to automatically add the --store optional and set up sandbox-paths correctly
<lopsided98>
This works perfectly except for when I tried to use the machines as hydra builders
<adamantium>
Hey guys, i enabled unfree and installed "google-chrome-dev" and in KDE there is no icon.
<lopsided98>
When Hydra runs builds, they are unable to find /bin/sh
<pie_>
lopsided98, neat
<gchristensen>
illegalprime: hey, I can't really help right now, but could you explain more about what is happening?
<lopsided98>
I can build the exact same derivation on the machine using "nix build" and it works fine
<illegalprime>
gchristensen: thanks & sure! I'm trying to build hello with a cross gcc 4.8 toolchain, when I run the build it attempts to build gcc-debug-4.8.5-armv7l-unknown-linux-gnueabihf but that build fails
<illegalprime>
I see gcc 7.4 paths when I'm building gcc 4.8, I think it's failing because of that
<illegalprime>
this path is getting included is /nix/store/9zqqkgaj9yw5l16c42578ykbwhc0py7h-armv7l-unknown-linux-gnueabihf-stage-final-gcc-debug-7.4.0/lib/gcc/armv7l-unknown-linux-gnueabihf/7.4.0/include-fixed
Rusty1 has joined #nixos
<lopsided98>
I upgraded from Nix 2.2 (which matthewbauer's static-nix branch has) to 2.2.2 and I thought it fixed it because I was able to build some things that previously failed, but now it has started happening again
talqu has quit [Remote host closed the connection]
mmlb39491 has quit [Ping timeout: 246 seconds]
<lopsided98>
AFAICT Hydra just runs "nix-store --serve --write", but I haven't figured out any way to reproduce the problem outside of Hydra
justanotheruser has joined #nixos
shibboleth has quit [Quit: shibboleth]
Anton-Latukha has quit [Ping timeout: 248 seconds]
kl3 has quit [Ping timeout: 246 seconds]
<illegalprime>
i think i might have to replace gcc with gcc48 globally but ive been unable to do that with an override, this is a similar issue: https://github.com/NixOS/nixpkgs/issues/38552
<{^_^}>
#38552 (by veprbl, 1 year ago, open): Override the default compiler
<infinisil>
illegalprime: I want to point out that you didn't even post your error.. Which is like the most important thing for trying to fix things
<matthewbauer>
illegalprime: i think you need: `nix-build -E 'with (import <nixpkgs> {}).pkgsCross.armv7l-hf-multiplatform; hello.override { stdenv = overrideCC stdenv buildPackages.gcc48; }`
Ariakenom has quit [Quit: Leaving]
Copenhagen_Bram has joined #nixos
<illegalprime>
infinisil: you're right, sorry about that. I'm trying to organize the relevant lines in a pastebin, it's kind of messy
<illegalprime>
matthewbauer: oh dang i think that worked, thanks! I'm trying to build hardkernel's uboot for the odroid c1+
<illegalprime>
I have to spend more time reading the internals of stdenv to grok it more
<illegalprime>
matthewbauer: the command you gave finishes but a `nix-store -qR result | grep -i gcc` gives gcc7 dependencies, not gcc48
<adamantium>
Question: running nixos-unstable branch with zfs... Should I worry that the nixos kernel might upgrade to 5.0 before zfs supports it?
<matthewbauer>
illegalprime: looks like it's using both. i wonder if it's getting the gcc-7 part from glibc? there is a way to override that but it's not as easy
shibboleth has joined #nixos
<infinisil>
adamantium: i think we have some assertions that would give you an error if that happened. And zfs already supports 5.0
<adamantium>
oh, great. thanks infinisil
<adamantium>
Glad you guys are on top of this stuff
<adamantium>
Is there a simple way to just pull in all of the kde applications
<adamantium>
I don't want to maintain a list, but I want everything to just be there
<illegalprime>
adamantium: I guess you could loop through everything in kdeApplications.*
<lopsided98>
I mostly figured out my problem with the sandbox /bin/sh. It is using a version of ld-linux that is not included in the sandbox (which results in the misleading "No such file or directory")
<illegalprime>
something like `builtins.attrValues pkgs.kdeApplications`
<lopsided98>
It works sometimes because the nixpkgs revision used to build the sandbox bash happens to match the current build
silver has quit [Quit: rakede]
<lopsided98>
Using a statically linked bash works correctly
<adamantium>
lopsided98: Ah, I was hoping someone would have done this already and just give me some code :) I don't know nix
<adamantium>
illegalprime: ^
Yaniel has quit [Ping timeout: 246 seconds]
<illegalprime>
adamantium: so you could try `environment.systemPackages = builtins.attrValues pkgs.kdeApplications;`
acarrico has joined #nixos
<illegalprime>
`builtins.attrValues pkgs.kdeApplications` returns a list of everything in kdeApplications, if you want to append it to an existing list you can do `[ pkg1 pkg2 ] ++ (builtins.attrValues pkgs.kdeApplications)`
<illegalprime>
adamantium: I would keep a list of packages though, the list in kdeApplications is really _a lot of_ packages, and you probably don't need them all
<adamantium>
error: The option value `environment.systemPackages.[definition 6-entry 13]' in `/nix-config/modules/nixos/desktop/plasma.nix' is not of type `package'.
erictapen has quit [Ping timeout: 245 seconds]
<adamantium>
illegalprime: Alright, i guess you're probably right. I'm always overkill on stuff
<bohan>
i notived gcc8 is actually gcc7
<bohan>
i need 9 anyway, heh
erictapen has joined #nixos
<illegalprime>
adamantium: oh i see, some values in kdeApplications are not packages, but are helpers or data, etc. you would filter that list to get just the packages. but again I don't think it's commonly done
<adamantium>
Yeah forget it. Thanks anyway
<adamantium>
A lot of people ask questions on how to do this online though via reddit, discourse, and github issues. We need documentation on how to do it I think
<adamantium>
Because none of the solutions on those pages really works
<illegalprime>
but yea if it's a popular question it should probably be in docs or FAQs
<bohan>
actually, i don't understand what's going on with gcc8. When i do nix-shell -p gcc8, it's putting /nix/store/rh9a7a202vcrknh3iv6xgxzqgnb4jadz-gcc-wrapper-8.3.0/bin in my PATH env var, but *after* another path to gcc-wrapper-7
<infinisil>
bohan: nix-shell -p implicitly uses stdenv.mkDerivation, giving all -p arguments via buildInputs