<infinisil>
Unrelated to the discussion itself happening in https://github.com/NixOS/nixpkgs/pull/56227, the fact that oxij's responses are really long all the time, incorrectly formatted (probably due email?), and they don't use github's comments on code feature, makes it *really* hard to follow the discussion..
<{^_^}>
#56227 (by oxij, 1 day ago, open): Typed `nixpkgs.config` with Gentoo-like use-flags
<drakonis>
ah this is nice.
<gchristensen>
infinisil: they are so long. I try to make my replies as brief as I can to not cause further reply growth
<infinisil>
I should aim for the same
<drakonis>
ah this is the SLNOS guy
drakonis has quit [Quit: WeeChat 2.3]
orivej has joined #nixos-dev
pie___ has joined #nixos-dev
pie___ has quit [Remote host closed the connection]
pie___ has joined #nixos-dev
pie__ has quit [Ping timeout: 268 seconds]
<ekleog>
infinisil: incorrect formatting is due to email indeed, I've poked GitHub already to ask them to at least add a checkbox to enable markdown formatting on emails, without any answer… feel free to poke them too, hopefully they'll end up adding it :)
<domenkozar>
so now that travis will die
<domenkozar>
what's the next go-to CI for nix?
<etu>
what, travis will die? O_O
<etu>
I heard that they got bought and a bunch laid of or something like that
<domenkozar>
they sold to a fund that is good at making existing customers happy and milk them
<domenkozar>
but you can be sure that OSS part will die as there's no direct revenue
<domenkozar>
just a question of time
<simpson>
It's good to have a plan, at least. Hope is not a strategy.
<domenkozar>
best case service will get worse close to no support, which was already at pain levels :)
<domenkozar>
anyway, what simpson said
<domenkozar>
circleci is quite ok
<domenkozar>
but we'd need to pimp up the docker image
<domenkozar>
something to make garbas happy
johanot has joined #nixos-dev
ckauhaus has joined #nixos-dev
ixxie has joined #nixos-dev
<garbas>
domenkozar: good luck, its sure possible, but somebody needs to have time for 2/3 days to really polish it out (eg, all the nix commands working in multi user installation)
<domenkozar>
it seems like there's no other choice
<Profpatsch>
domenkozar: What about ofborg?
<Profpatsch>
Most of what a good github CI does is already there.
<Profpatsch>
It’s specific to nixpkgs at the moment, but that can be solved by abstracting out the specific parts.
<Mic92>
Is running VMs build with nixos-rebuild build-vm broken atm?
orivej has quit [Ping timeout: 258 seconds]
<zimbatm>
with the `machine: {enabled: true}` attribute on a job it's possible to get a multi-user setup on CircleCI
<zimbatm>
with sandboxing enabled
<zimbatm>
the main thing missing is kvm support for Qemu tests
<thoughtpolice>
So, just out of curiousity -- are 'nix copy --to 's3://...' and 'nix copy --to file://... && <upload directory to S3>' functionally equivalent? I get the impression they are, although the first obviously doesn't write to disk.
<LnL>
you can nix copy --from file:// --to s3://
<thoughtpolice>
Yes but then I need to keep the list of paths/roots around, don't I?
<thoughtpolice>
<INSTALLABLES> are required for nix copy (which IMO is a bug, but anyway...)
<LnL>
there's an --all flag
<thoughtpolice>
Oh! 💯
<LnL>
pretty sure that will diff with the s3 store and upload what's missing
<thoughtpolice>
Yeah, though I think most tools can do that. The --all thing is really the only reason I didn't use 'nix copy' directly anyway
Melkor333 has quit [Quit: WeeChat 2.4]
init_6 has quit [Read error: Connection reset by peer]
init_6 has joined #nixos-dev
emily has quit [Remote host closed the connection]
emily has joined #nixos-dev
init_6 has quit []
init_6 has joined #nixos-dev
Melkor333 has joined #nixos-dev
init_6 has quit []
emily has quit [Remote host closed the connection]
<Profpatsch>
I had a horrified realization right now: --show-trace is not documented in any manpage.
<xeji>
rename it to --ghost-trace :)
<clever>
:O
<samueldr>
(I'll be the one who well actually's)
<samueldr>
well, actually, it's in nixos-rebuild's
<samueldr>
:)
<Profpatsch>
lol
<samueldr>
not that it's useful, considering nixos-rebuild isn't in nix
<tilpner>
Anyone familiar with buildPythonPackage? I get "Could not find a version that satisfies the requirement logbook (from matrix-nio==0.1) (from versions: )" even though python3Packages.Logbook is in propagatedBuildInputs
johanot has quit [Quit: WeeChat 2.2]
MichaelRaskin has joined #nixos-dev
kisik21 has joined #nixos-dev
tilpner has quit [Quit: WeeChat 2.4]
<samueldr>
I'm thinking maybe an informal, and informational, page for "erratas" (known issues?) between channels might be something to look into figuring out
<samueldr>
I'm thinking things like between 18.09 and 19.03/unstable how gdk_pixbuf-using applications are causing issues
orivej has quit [Ping timeout: 240 seconds]
<samueldr>
not sure what and how exactly, which is why I'm tossing the idea here
orivej has joined #nixos-dev
<MichaelRaskin>
Do you mean regressions?
kisik21 has quit [Remote host closed the connection]
<samueldr>
not sure about the right name :)
<samueldr>
but it's not forcibly a _regression_ rather an incompatibility in that case
<MichaelRaskin>
How should it compare with release notes?
<samueldr>
18,09 release notes are compared to the past
<samueldr>
18.09 "$name" are about future versions
orivej has quit [Ping timeout: 240 seconds]
<samueldr>
(good questions, helps untangling the nascent idea in my mind)
<MichaelRaskin>
Is "$a" "$b" just repease notes for "$b" and previous releases, up to "$a" (exclusively)?
<MichaelRaskin>
Test-driven development!
<MichaelRaskin>
«If it is different, it should pass the test of ‘‘it is just …’’ by providing a difference»
<samueldr>
not sure I follow here
<MichaelRaskin>
Well, in test driven development you don't really know what you are writing, but slowly accumulate tests «in any case it should do something sane in this case»
<MichaelRaskin>
Here we are not sure what you are describing, but probably there should be an explainable difference from release notes
<samueldr>
>> Is "$a" "$b" just [release] notes for "$b" and previous releases, up to "$a" (exclusively)?
<samueldr>
ah
<samueldr>
I must have been confused in my initial reading
kisik21 has joined #nixos-dev
<samueldr>
release notes for release $b are differences only between "$a" and "$b" (but this probably also makes them differences from all previous ancestors)
<samueldr>
at least, I think so
<samueldr>
I don't think we have precedent of adding notes to the release notes after release. I don't know if in itself this should or shouldn't happen, but I'm thinking it should always be relative to previous releases
<MichaelRaskin>
We do start writing next-release release notes more or less right after the release
Melkor333 has quit [Quit: WeeChat 2.4]
<MichaelRaskin>
I would hope that the union of adjacent-release differences is a good approximation of distant-release differences
<samueldr>
the update to gdk_pixbuf (let's pick on that example again) isn't release note worthy; it isn't itself "a note"
<samueldr>
but the combination of an older gdk_pixbuf with software using the newer one creates an incompatibility
<samueldr>
I'm not sure if there's anything more to do than have the issue, for "well known" incompatibilities
<MichaelRaskin>
Hm, so it is about things specific for multichannel setups?
<samueldr>
I think it'd be the most likely thing to find there
<samueldr>
but maybe we'd put things like the kernel regression we just had
<samueldr>
more about "oopsies" than "this changes", if you allow the infantilizing description
<MichaelRaskin>
Well, we have a ton of open issues that don't always get a constructive reaction (just too many of them), so it might be hard to draw a cutoff for inclusion
kisik21 has quit [Read error: Connection reset by peer]
tilpner has joined #nixos-dev
tilpner has quit [Remote host closed the connection]