srk has quit [Read error: Connection reset by peer]
srk has joined #nixos-dev
orivej has quit [Ping timeout: 264 seconds]
supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-dev
cole-h has joined #nixos-dev
rajivr has joined #nixos-dev
<supersandro2000>
How do I resolve a circular dependency in python packages?
<hexa->
usually happens during tests, disable them on one end.
<supersandro2000>
makes sense. thanks.
<supersandro2000>
I think I will go the pytest hypothesis way and disable the test for one dependency
<supersandro2000>
seems like a common problem around that lib
<supersandro2000>
"lets quickly update that one dependency and then only one is missing for that package for python 3.9"
<supersandro2000>
me rebuilding pytest because a dependency of it needed an update because the update to the lib draws in a new dependency which required that 🤯
red[evilred] has quit [Quit: Idle timeout reached: 10800s]
tokudan has quit [Remote host closed the connection]
<worldofpeace>
Hey, is there anyone here willing to review some qemu PRs, or who usually does?
<qyliss>
I can review QEMU PRs
<qyliss>
I should add myself as a maintainer
<qyliss>
worldofpeace: can't easily access my GH account for the next few days, but I'm happy to look over them / test them and let you know if they're good to merge in the meantime
<qyliss>
(although I do realise that some of the PRs you're talking about are probably mine :P)
<qyliss>
a lot of the time when you're running a NixOS VM you _want_ it to ignore filesystems because you don't want it to try to mount things that don't exist
<worldofpeace>
qyliss: perhaps qemu should have it's own filessytems option?
<worldofpeace>
* qyliss: perhaps qemu should have its own filessytems option?
<qyliss>
Yeah I'd say so
<qyliss>
At least to start with
mkg20001 has joined #nixos-dev
<mkg20001>
#112746 is about qemu getting it's own filesystems option... virtualisation.fileSystems isn't fileSystems
<qyliss>
mkg20001: yes, but it shouldn't default to options.fileSystems
<mkg20001>
afaik config.fileSystems contains the filesystems and options.fileSystems simply contains the layout of the option, meaning you can set values like in fileSystems but it doesn't copy the contents of it
<worldofpeace>
hmm I'm seeing a lot of nixos tests with fileSystems = pkgs.lib.mkVMOverride
<qyliss>
mkg20001: ah whoops sorry I misread
<qyliss>
yeah in that case #112746 LGTM although I haven't tested it
<qyliss>
it would be nice if all those tests were updated to use it (would prove it works very well!) but that shouldn't block getting it merged
<worldofpeace>
yeah that would be cool
<mkg20001>
I already did the linked one to test
<mkg20001>
The ones without also works
<mkg20001>
What's the commit msg for updating a test?
<worldofpeace>
you can build all tests like nixosTests.gnome3
<worldofpeace>
corresponding to the attribute in I think all-tests.nix
<mkg20001>
pushed a commit to update all testss to virtualisation.fileSystems
<mkg20001>
running the changes ones so far looks good
<mkg20001>
qyliss: can I merge the PR?
<qyliss>
lgtm
orivej has joined #nixos-dev
justanotheruser has quit [Ping timeout: 260 seconds]
justanotheruser has joined #nixos-dev
<gchristensen>
would anyone be opposed to me backporting all the tailscale updates in unstable to stable? it brings in some features that are otherwise unavailable (magic dns)
<qyliss>
Where are the packages required for nixos-unstable-small defined?
<flokli>
gchristensen: tell me more about that magic ;-)
<rmcgibbo[m]>
Well, luckily that one wouldn't actually trigger because the outer construct is a List.
<rmcgibbo[m]>
It only falls to that branch of trawling the value for lists if the value that's assigned to the `patches` key isn't itself a list.
<jtojnar>
rmcgibbo: aactually, fetchpack takes some list arguments (e.g. excludes)
<jtojnar>
*fetchpatch
<rmcgibbo[m]>
oh oh oh... you're totally right, i think that dooms it.
<jtojnar>
maybe only match lists that are not inside another list?
<jtojnar>
that still adds false negatives but no false positives
<jtojnar>
AFAICS
<jtojnar>
(stopping the traversal in the patches value node, of course)
<rmcgibbo[m]>
`patches = lib.singleton substituteAll { src = ...; path = makeBinPath [ foo bar]; }` would still be a false positive in that scheme
<jtojnar>
hmm, right
<rmcgibbo[m]>
or `lib.singleton fetchpatch {... excludes = [foo bar]}` is probably more likely in practice
<rmcgibbo[m]>
all of these are pretty rare though
<Gaelan>
Anyone able to help with some git-fu? I'm trying to fix the merge conflict in 110571 (a staging PR), but I can't seem to do it without github pulling in a bunch of random PRs