jushur has quit [(Quit: The Borg joined forces with Skynet, Resistance is futile! Uploading has begun!)]
jushur has joined joined #nixos-dev
Sonarpulse has quit [(Ping timeout: 240 seconds)]
<gchristensen>
I'm starting to feel the PR volume is a problem. I didn't think it was when we were steady-state at <300 but 485? :/
<copumpkin>
Mic92: if you have other ideas to try
<copumpkin>
I'm happy to try :)
<copumpkin>
I've spent a lot of time staring at the outputs of the -X commands, and adding prints to the rustc source code and recompiling it, and various other crap
<copumpkin>
most of what I found is on the github tickets, but I don't really know enough about rustc to go further
<Mic92>
copumpkin: what linker have you tried?
<copumpkin>
the official apple one from outside of nix
mbrgm has quit [(Ping timeout: 255 seconds)]
<copumpkin>
but I probably need to rebuild rustc with it come to think about it
<copumpkin>
since the issue is that rustc thinks the stdlib is b0rked
mbrgm has joined joined #nixos-dev
<Mic92>
copumpkin: it is probably faster to iterate to build rust from git in a nix-shell
<Mic92>
copumpkin: ar is also not a terrible complex format
<copumpkin>
yeah I know
<copumpkin>
so the official llvm tooling says the same thing about the rust libraries
<copumpkin>
llvm-nm, etc.
<Mic92>
copumpkin: you can also install the rust std-lib with rustup and check if rustc from nix can use it correctly
infinisil has quit [(Quit: ZNC 1.6.5 - http://znc.in)]
Jackneilll has quit [(Quit: Leaving)]
<ekleog>
gchristensen copumpkin: oh thanks! :)
Jackneill has joined joined #nixos-dev
<gchristensen>
dammit
<gchristensen>
I'm sorry
<gchristensen>
ok, the patches are out to 17.09
<gchristensen>
niksnut: ping?
<niksnut>
gchristensen: pong
<gchristensen>
I have a Type-2 provisioning, can you update Hydra when it is ready? this time the / FS isn't ZFS, so I should be able to recover it if something goes wrong
<niksnut>
sure
<gchristensen>
ok, great
<gchristensen>
I wonder if I could even just netboot it, and substitute the SSH host key, that way some mystery failure in the future doesn't require a whole process to fix it
<gchristensen>
but that is a thing to try later :P
<gchristensen>
niksnut: packet-t2-4 at 147.75.98.145 will be ready shortly
<gchristensen>
oh, should be ready now :)
<fpletz>
gchristensen: awesome :)
<gchristensen>
I'm sorry it took so long, I wanted to get ZFS out of the base image so next time it failed I had a chance to fix it
<gchristensen>
hydra feels like it goes in slow motion without it heh
<gchristensen>
fpletz: do you have any favorite hosted host monitoring tools? I'd love to get more data off of them but without too much work
<gchristensen>
or cost :)
<fpletz>
gchristensen: not really, sorry. I'm a fan of self-hosted solutions and am currently using prometheus/alertmanager for most stuff
<gchristensen>
aye :)
<domenkozar>
gchristensen: datadog
<domenkozar>
I'm a fan of hosted solutions since I can then enjoy my life :)
<gchristensen>
hey now :)
<domenkozar>
datadog is pretty cool, but it's not cheap
<gchristensen>
datadog only does 1d retention for free, which is a bit low
<domenkozar>
also depends what you need
<domenkozar>
it's not easy to be percentiles
<gchristensen>
I'd settle for like 3 day retention
<domenkozar>
to do*
<domenkozar>
yeah they have restrictive freemium account
<domenkozar>
I guess they want you to pay :)
<gchristensen>
maybe I'll send them a nicely worded email asking for a free 1host account
<gchristensen>
mm... 2 host account
<domenkozar>
right
<domenkozar>
good luck :)
<copumpkin>
what are their rates for longer retention?
<gchristensen>
$15/mo/host
<copumpkin>
when they say full-resolution data
<copumpkin>
does that mean full log retention?
<gchristensen>
yeah, I think that means they don't aggregate any data to save space, but keep the full original reports
<copumpkin>
ah
<gchristensen>
niksnut: bump on that hydra node
<niksnut>
thanks
<niksnut>
how many cores?
<gchristensen>
same as the other type-2
<niksnut>
ok
<domenkozar>
peti: question: does hackage-db allow specifying what snapshot to use?
<domenkozar>
by timestamp or something else
phreedom has quit [(Ping timeout: 240 seconds)]
<peti>
domenkozar: Yes. readTarball takes an optional UTCTime argument for that purpose
<domenkozar>
peti: any objections exposing that via cabal2nix cli?
phreedom has joined joined #nixos-dev
phreedom has quit [(Quit: No Ping reply in 180 seconds.)]
phreedom has joined joined #nixos-dev
<peti>
domenkozar: no objections, sounds like a good idea.
<niksnut>
gchristensen: okay, it seems to be working
<gchristensen>
great :)
<niksnut>
I guess I can remove packet-t2-3?
<gchristensen>
yeah
<gchristensen>
:(
<niksnut>
ok, thanks for the new machine :-)
<gchristensen>
:) I'm glad to get hydra back to full capacity
<copumpkin>
o/
<niksnut>
copumpkin: hopefully something along the lines of nix build --dry-run --store local?root=/tmp/dummy --substituters https://... works
<copumpkin>
yeah I was playing with the ?root=/tmp/dummy thing
<copumpkin>
but it seems to behave differently w.r.t. my one IFD for pulling in nixpkgs
<copumpkin>
I guess I could make that behavior conditional
<niksnut>
I'm constantly struggling with the right behaviour for --dry-run
<copumpkin>
also, if I pass in a ?root=, do I also need to pass in ?store=/nix/store to convince it that my store dir is unchanged?
<niksnut>
it would be so nice to get rid of .drvs
<niksnut>
then evaluation wouldn't need a writable store
<copumpkin>
yeah, it seems tricky when it comes to IFD
<niksnut>
or, at least, writing .drvs could be delayed
<copumpkin>
yeah
<copumpkin>
you had a plan for no-.drv builds didn't you?
<niksnut>
so the result of evaluation would be a graph of files to be written to the store
<niksnut>
I wouldn't call it a plan
<niksnut>
:-)
<copumpkin>
yeah
<niksnut>
more a vague hope
<copumpkin>
nothing forces them into the store in the first place, right? eval could be a "derivation" whose output contains a bunch of .drv files and gets realized by copying into the store
<copumpkin>
it seems like the store is just a convenient place to put them
<copumpkin>
but they could be ephemeral or only moved to the store later
__Sander__ has quit [(Quit: Konversation terminated!)]
<copumpkin>
mv: cannot move '/private/var/folders/6j/566zczg137j_zbvhjh2zcpr80000gn/T/nix-build-nixpkgs-90e78aafadba43f9ec1190d9411720d561f1e09b-src.drv-0/unpack/nixpkgs-90e78aafadba43f9ec1190d9411720d561f1e09b' to '/nix/store/lfcki0xb6bd0c3cx223qnflwdcryk1ik-nixpkgs-90e78aafadba43f9ec1190d9411720d561f1e09b-src/nixpkgs-90e78aafadba43f9ec1190d9411720d561f1e09b': Permission denied
<copumpkin>
unclear why it's trying to write to /nix/store given my NIX_REMOTE
<copumpkin>
oh, possibly because I'm building in a different store
<copumpkin>
$ nix copy --to local?root=/private/tmp/dummy-store?require-sigs=false $(nix-build default.nix -A pkgsBundle)
<copumpkin>
[0 copied (50.7 MiB)]
<copumpkin>
error: cannot add path '/nix/store/lfcki0xb6bd0c3cx223qnflwdcryk1ik-nixpkgs-90e78aafadba43f9ec1190d9411720d561f1e09b-src' because it lacks a valid signature
<niksnut>
I don't think that whole root= feature works on macOS
<copumpkin>
it did!
<copumpkin>
or seems to be :)
<copumpkin>
why wouldn't you expect it to?
<copumpkin>
niksnut: is there a way to ask that a particular nar not be compressed with nix copy? :) :)
<niksnut>
no
Sonarpulse has quit [(Ping timeout: 240 seconds)]
romildo has joined joined #nixos-dev
JosW has quit [(Quit: Konversation terminated!)]
<fpletz>
gchristensen: I've just set up my first packet.net box… is the code that generates the packet.nix files somewhere? the ipv6 default gateway is missing for example… and we need a 17.09 template :)
<gchristensen>
oh?
<gchristensen>
17.09 is almost ready!
<gchristensen>
I have a set of new images almost ready to send over to the team
<gchristensen>
I'm also switching all the instance types away from ZFS (as noted in the commits)
<gchristensen>
what kind of instance did you get?
<copumpkin>
omg no zfs
* copumpkin
faints
<gchristensen>
it was really nice, but (1) the servers kept just falling over for some reason (2) the rescue OS made it hard-to-impossible to mount the drives and figure out what happened
<copumpkin>
aw
<copumpkin>
oh well, maybe someday :)
<Mic92>
do we have more pull requests than in the past?
<copumpkin>
yeah
<copumpkin>
I wonder if there are charts over time anywhere in github
<gchristensen>
more pull requests open? or in rate?
<Mic92>
rate
<gchristensen>
ah
<Mic92>
open vs close rate would be also interesting to me.
<gchristensen>
IIRC I had a program to calculate this, maybe I can find it
<domenkozar>
and everyone recommends using ZFS, good to hear real world usage :)
<gchristensen>
I use ZFS on several machines, it works great
<gchristensen>
its different on a remote server without adequate rescue tools :P
nixer has quit [(Ping timeout: 260 seconds)]
<domenkozar>
gchristensen: for me all servers are remote
<gchristensen>
without adequate rescue tools?
<gchristensen>
I use it on remote servers too, but with better rescue tools :)
<domenkozar>
I guess I don't understand why you can't have the tools
<gchristensen>
with Packet all you can do is reboot in to Alpine, which is a terrible pain to get ZFS working
<domenkozar>
it should allow you to install nix, or is it read-only fs?
<gchristensen>
you can't specify arbitrary disk images to boot or a one-time netboot thing, or unmount the disk from your virtual system and remount to another virtual system
<Mic92>
gchristensen: you can use clever's kexec image :)
<gchristensen>
it isn't entirely read-only, but certainly some of it
<gchristensen>
and you can't use kexec, because the grsec kernel has kexec disabled
<gchristensen>
and everything you do has to fit in memory
mingc has quit [(Quit: Ping timeout (120 seconds))]
Sonarpulse has joined joined #nixos-dev
MichaelRaskin has joined joined #nixos-dev
pbogdan has quit [(Quit: ZNC 1.6.5 - http://znc.in)]
pbogdan has joined joined #nixos-dev
<copumpkin>
trying to write idiomatic Nix C++. What's the proper C++ way to malloc a buffer (for a C syscall) such that the destructor takes care of freeing it?
<Dezgeg>
vector<char> or std::string I suppose?
<copumpkin>
well I need a legit pointer that the syscall can write to
<copumpkin>
I thought vector didn't expose that
<Dezgeg>
data()
<copumpkin>
hmm okay, that might make sense then
<copumpkin>
it's actually conceptually an array of structs
<copumpkin>
if that helps
<copumpkin>
I guess I could make it vector<mystruct>?
<Dezgeg>
yes
<copumpkin>
that might be nicer than I was hoping for then :D
<Dezgeg>
oh, unique_ptr<foo[]> also would work... TIL
<copumpkin>
ah
<copumpkin>
vector is probably nicer for other reasons