justanotheruser has quit [Ping timeout: 246 seconds]
justanotheruser has joined #nixos-dev
justanotheruser has quit [Client Quit]
lovesegfault has quit [Ping timeout: 268 seconds]
drakonis has quit [Quit: WeeChat 2.6]
das_j has quit [Remote host closed the connection]
Scriptkiddi has quit [Remote host closed the connection]
Scriptkiddi has joined #nixos-dev
das_j has joined #nixos-dev
orivej has joined #nixos-dev
<genesis>
since we have fromJson and toJSON, i wonder if we can write some Json schema to describe desktop entry, then use a json tool as https://github.com/any-json/any-json to parse them
<genesis>
describe exotic format with json schema would lead to validation and easier maintenable code (or not).
justanotheruser has joined #nixos-dev
orivej has quit [Ping timeout: 268 seconds]
lovesegfault has joined #nixos-dev
ivan has left #nixos-dev [#nixos-dev]
pie_ has quit [Ping timeout: 268 seconds]
__monty__ has joined #nixos-dev
__Sander__ has joined #nixos-dev
_ris has quit [Ping timeout: 258 seconds]
FRidh has quit [Quit: Konversation terminated!]
<globin_>
worldofpeace, niksnut, domenkozar[m], shlevy, Mic92, sphalerite, spacekookie, I won't be able to make it today, sorry
<worldofpeace>
ooh, did I miss a meeting this time?
<worldofpeace>
I was just thinking about this yesterday and I was like, hmm wasn't that sorta going to happen next week since the nomination
<worldofpeace>
in the future, If i have a date I kind try to make things in happen. just an advance on the info and it's more likely to happen. I'm EST
__Sander__ has quit [Quit: Konversation terminated!]
<gchristensen>
make *everybody*'s :) jtojnar did incredible work just yesterday removing python2 from like 50 packages
<gchristensen>
(that is what motivated me to look at my own system)
<worldofpeace>
jtojnar++
<{^_^}>
jtojnar's karma got increased to 22
<worldofpeace>
jtojnar is a pretty cool and motivating person, IMHO :D
<Taneb>
One of the packages I maintain needs a) updating and b) switching to Python 3... I'll try to do that tonight (r-ryantm made a PR but I never quite got round to it)
<Taneb>
Should nix-build "<nixpkgs>" -A python be changed so it's Python 3?
<gchristensen>
(tbh I'm not sure actually if this flame graph view is useful or not...)
<Taneb>
gchristensen: I don't have a clear idea of what the x-axis is. "how many ways does this depend on python2?"
<gchristensen>
yeah
<domenkozar[m]>
I have nixops in my closure so I'll just have a beer tonight :)
<LnL>
:p
<gchristensen>
ehh my output is buggy.
<LnL>
the darwin stdenv also still uses it at the moment
<gchristensen>
yarr, bit by caching
orivej has joined #nixos-dev
<Taneb>
The Wikipedia policy of "be bold" applies to nixpkgs PRs, right?
<worldofpeace>
erm, Taneb not familiar. are you speaking of our wiki policy/s?
<worldofpeace>
erm, there's some things I'm not completely alligned with there
<worldofpeace>
But I take PRs in an exercise of grace and dignity
<worldofpeace>
and understand
<worldofpeace>
-ing
<Taneb>
Really, I'm asking "should I make a really bold PR and have discussion after making it, or should I try to discuss the change before making the PR"
<worldofpeace>
Like I know a process of creating an issue before creating a purposal, but some discussion has to occur for the change to be accepted anyway. For example with the Python port for testing Pr, you could say this was a bold change but discussions happened offline that shaped it into being a **bolder** proposal. I'm one to be say yes && no, and depending on if you think the change is one of difficulty I wouldn't reccomend
<worldofpeace>
investing your time without creating some sort of connection here with what is actually wanted.
<worldofpeace>
So I think either way you could get what you want.
<worldofpeace>
Personally, I think people in nixpkgs really become invested if they see a certain quality in a proposal.
<worldofpeace>
So I don't advise contributing in a way where you create a PR but you're disconnected from the whole giving the change.
<drakonis>
there's another fork that will maintain python post EOL
<emily>
i personally think that would be good but there are backcompat argumets not to do it
<emily>
niksnut: 2017 :'(
<emily>
drakonis: i consider all such forks speculative until they actually fix a real security-relevant py2.7 issue in a timely manner tbh
<niksnut>
I think changing python -> python3 has the same issue, it will silently break programs
<emily>
RHEL is the only one with money on the line to actually do that
<drakonis>
yes
<niksnut>
better if anything to remove the python attribute
<drakonis>
fair enough
<infinisil>
I'm fro dropping python 2 completely, nobody should use it anymore, and if they really need to there's older nixpkgs versions
<worldofpeace>
In short Taneb I believe you catalyzied something important. What's happening in nixpkgs for python2 is not clear for people who want to help
<niksnut>
I disagree with dropping python 2, that's just not realistic and also not necessary
<niksnut>
software doesn't stop working just because it has been EOLed
<emily>
it does start rotting thuogh
<niksnut>
and the amount of software using python 2 is enormous
<emily>
*though
<worldofpeace>
niksnut++
<{^_^}>
niksnut's karma got increased to 13
<emily>
the amount of python 2 software in nixpkgs is enormous
<emily>
much of it works with python 3 and just hasn't been updated
<niksnut>
that too :-)
<emily>
at least switching to pypy would force everyone to try upgrading to py3 the stuff that broke :p
<worldofpeace>
though there's cases where it's a very good idea to ditch things
<worldofpeace>
if you care about your maintainers
<emily>
*someone* has to lead the py2 deprecation, major libraries and applications are already there
<emily>
it's been 11 years, waiting longer isn't going to do anything :/
<niksnut>
we still need to rewrite nixops to python 3...
<drakonis>
oh, that's a problem.
<emily>
lol fuck
<emily>
wonder whether it works in pypy
<drakonis>
time to test it out
<drakonis>
how complex can it be to port something from python 2 to 3?
<drakonis>
its probably the worst question to ever make
<worldofpeace>
lol
<worldofpeace>
drakonis: I enjoyed it though
<emily>
drakonis: apply judicious 2to3 then not too bad
<emily>
mostly you'll spent 5 months dealing with subtle unicode vs. bytes issues
<emily>
disclaimer: haven't actually done this for a large project
<drakonis>
unless you're leaning a lot on python 2 specific features
<drakonis>
i read a bit on the transition to py3 and it was pretty much growing pains
<drakonis>
not really a lot of stuff that needs significant changing
<emily>
it's certainly not so hard a transition that it isn't a bit embarrassing that nixops is still on 2 (but I'm not volunteering to do the work, so *shrug*)
<Taneb>
There's a little note in nixops saying "we can do this when python 2.7.13 comes out"
<simpson>
drakonis: It is far easier to go from CPython to CPython+PyPy compatibility, than to do Python 2 to Python 3. They're really different languages.
<emily>
Taneb: lol
<emily>
simpson: except not really, since the big "FFI"-y/C extension libraries are available on Python 3 and frequently not on PyPy
<emily>
like, on a linguistic level, yeah
<emily>
in practice given where the Python community is at, I don't think that's really true in practice
<simpson>
emily: Ah, sure. I think that most of the low-hanging fruit's already ported, too; the main holdouts are libraries like lxml where there's active upstream resistance to PyPy.
<simpson>
emily: Specifically, lxml doubts cffi, leading to this hilarity: https://github.com/lxml-cffi/lxml-cffi Such is life. I kind of agree with niksnut; let's remove the `pythonPackages` name entirely, as it is too silly.
<emily>
I'd be +1 for making it error or at least warn
lovesegfault has quit [Ping timeout: 252 seconds]
<Taneb>
I had a quick go at upgrading nixops to use python3
<Taneb>
There's one test that seems to rely on sets preserving order which I think was true in python2 but not python3
<qyliss>
What to we need to do to ban buildRustPackage from Nixpkgs (and replace with e.g. crate2nix)?
<qyliss>
Feels like it might be a little small to be an RFC, so wondering what other people think
<gchristensen>
I think an RFC could be good
<gchristensen>
as a mechanism of information sharing
<gchristensen>
and precedent setting for future output-hashed data
drakonis has quit [Ping timeout: 246 seconds]
<qyliss>
That's the answer I was afraid of, but think you might be right
<gchristensen>
it might be worth collaborating with Eelco on it, he and I were talking about this a while ago, thinking about what is the litmus test for fixed output builds
<qyliss>
I'm not sure I want to get too deep into what is and is not acceptable right now
<gchristensen>
ok
<qyliss>
I worry that would hold it up
<qyliss>
And I think buildRustPackage needs to be banned yesterday
<gchristensen>
+1
<qyliss>
But you could see it as a good first step
dtz has quit [*.net *.split]
<qyliss>
Towards properly defining that
<gchristensen>
sgtm
phreedom_ has quit [*.net *.split]
_ris has joined #nixos-dev
<qyliss>
I'd be happy to discuss what the criteria should be after, for sure.
drakonis has joined #nixos-dev
drakonis has quit [Quit: WeeChat 2.4]
drakonis1 has joined #nixos-dev
drakonis1 has quit [Client Quit]
drakonis has joined #nixos-dev
<Taneb>
PR exists on nixops to switch to python 3, that passes tests. I'd appreciate if someone who actually uses nixops could test I haven't broken anything
<andi->
qyliss: thanks, will try to give you feedback
<qyliss>
ty
<samueldr>
lovesegfault: what were the failures on aarch64?
<samueldr>
(hydra doesn't keep a log history for restarted jobs)
<lovesegfault>
samueldr: TImeout building the kernel
<samueldr>
ah
<lovesegfault>
which in turn caused everything to fail
<samueldr>
yeah
<andi->
qyliss: are you aware of naersk? I am thinking that it might ressemble the use case of buildRustPackage more closely but doesn't came with all the good benefits buildRustCrate gives us
<lovesegfault>
andi-: o/
<worldofpeace>
On the topic of Hydra, I believe I'd need a hydra account for 20.03 release mngmt. not sure who to contact.
<lovesegfault>
samueldr: is it possible to give priority to a job in Hydra?
<samueldr>
yes, but only few admins have that power
<qyliss>
andi-: what benefits are you thinking of?
<lovesegfault>
samueldr: Ah, I see
* lovesegfault
checks his powers
<lovesegfault>
I have restart powers
<lovesegfault>
Dope
<andi->
qyliss: well it works without generating additional files and just needs the lockfile (+ Cargo.toml). That might be a reasonable way if we do not want to keep generated files in nixpkgs.