evanjs has quit [Read error: Connection reset by peer]
evanjs has joined #nixos-dev
kini has quit [Ping timeout: 258 seconds]
<jtojnar>
domenkozar <3
<jtojnar>
"allows evaluating a flake per system" – is not the evaluation lazy anyway?
<domenkozar[m]>
that's bad wording, what I mean is: if you want to evaluate two systems, you evaluate the flake twice
<niksnut>
it says it's hermetic, but isn't, e.g. building the "foo" attribute of a flake can produce a different result for different users
kini has joined #nixos-dev
<niksnut>
so flake output attribute names no longer uniquely determine the result
<domenkozar[m]>
niksnut: what is the definition of hermetic?
<domenkozar[m]>
depending how you define it, it is or isn't
<jtojnar>
"Enumerating supported systems ... can differ between flake attributes" – the current state might be considered good if you have multiple packages with different set of supported platforms
<jtojnar>
but maybe it is good to have small flakes anyway
<domenkozar[m]>
niksnut: if you always evaluate flakes for all supportedSystems, it's hermetic by that definition
<jtojnar>
Is not the system implicit if you do `nix build .#attribute` too?
<niksnut>
at the UI level, yes
<niksnut>
but .#attribute gets expanded to e.g. .#builds.<system>.attribute
<niksnut>
and having a unique name is important
<domenkozar[m]>
so the same would happen now, except the system would be passed as an input
<niksnut>
e.g. it gives hydra a unique job name
<domenkozar[m]>
and <system> would be implicit by the input instead of the output
<domenkozar[m]>
while there could be a flag --local-system that would evaluate the flake with only current system instead of all
<domenkozar[m]>
or the other way around like --all-systems
<jtojnar>
maybe this can tie into the shlevy's URI proposal and we could have builds/attribute?system=foo as uri
<niksnut>
this also removes the possibility to depend on outputs for different system types, e.g. buildInputs = [ flake1.packages.i686-linux.foo flake2.packages.x86_64-linux.bar ]
<niksnut>
since 'system' is now a global thing
<domenkozar[m]>
there's nothing preventing one from mixing systems in the outputs
<domenkozar[m]>
it's the same as now, you can lie about what systems are in the system specific outputs
<niksnut>
how do you refer to flake outputs for a different system type?
<domenkozar[m]>
nix build --system x86_64-darwin
<niksnut>
I mean, how do you refer to different system types in the same eval? like in the buildInputs above
<domenkozar[m]>
oh, you can hardcode the system instead of relying on the inputs
<domenkozar[m]>
so: if system == 'x86_64-linux' then <evaluate nixpkgs with i686> else null
devhell has quit [Ping timeout: 246 seconds]
<domenkozar[m]>
that bit is still ugly as it was before and it's rarely needed
tomberek has quit [Quit: Connection closed]
devhell has joined #nixos-dev
<domenkozar[m]>
niksnut: ah, you're asking how to mix systems from different flakes?
<domenkozar[m]>
that would need to be supported by specifying stuff on the inputs side I guess
colemickens has quit [Quit: Bridge terminating on SIGTERM]
Ox4A6F has quit [Quit: Bridge terminating on SIGTERM]
Dandellion has quit [Quit: Bridge terminating on SIGTERM]
Valodim[m] has quit [Quit: Bridge terminating on SIGTERM]
zuh0 has quit [Quit: Bridge terminating on SIGTERM]
jtojnar has quit [Quit: Bridge terminating on SIGTERM]
symphorien[m] has quit [Quit: Bridge terminating on SIGTERM]
michaelpj has quit [Quit: Bridge terminating on SIGTERM]
rnhmjoj has quit [Quit: Bridge terminating on SIGTERM]
kalbasit[m] has quit [Quit: Bridge terminating on SIGTERM]
jonge[m] has quit [Quit: Bridge terminating on SIGTERM]
LinuxHackerman has quit [Quit: Bridge terminating on SIGTERM]
thefloweringash has quit [Quit: Bridge terminating on SIGTERM]
davidak[m] has quit [Quit: Bridge terminating on SIGTERM]
regnat[m] has quit [Quit: Bridge terminating on SIGTERM]
nh2[m] has quit [Quit: Bridge terminating on SIGTERM]
kloenk has quit [Quit: Bridge terminating on SIGTERM]
ryantm has quit [Quit: Bridge terminating on SIGTERM]
roberth has quit [Quit: Bridge terminating on SIGTERM]
domenkozar[m] has quit [Quit: Bridge terminating on SIGTERM]
Guest85450 has quit [Quit: Bridge terminating on SIGTERM]
aanderse has quit [Quit: Bridge terminating on SIGTERM]
danielrf[m] has quit [Quit: Bridge terminating on SIGTERM]
emily has quit [Quit: Bridge terminating on SIGTERM]
kraem[m] has quit [Quit: Bridge terminating on SIGTERM]
mjlbach has quit [Quit: Bridge terminating on SIGTERM]
siraben has quit [Quit: Bridge terminating on SIGTERM]
veleiro has quit [Quit: Bridge terminating on SIGTERM]
aaronjanse has quit [Quit: Bridge terminating on SIGTERM]
philipp[m] has quit [Quit: Bridge terminating on SIGTERM]
puzzlewolf has quit [Quit: Bridge terminating on SIGTERM]
zimbatm[m] has quit [Quit: Bridge terminating on SIGTERM]
rmcgibbo[m] has quit [Quit: Bridge terminating on SIGTERM]
garbas[m] has quit [Quit: Bridge terminating on SIGTERM]
JJJollyjim has quit [Quit: Bridge terminating on SIGTERM]
ma27[m] has quit [Quit: Bridge terminating on SIGTERM]
arcnmx has quit [Quit: Bridge terminating on SIGTERM]
DamienCassou has quit [Quit: Bridge terminating on SIGTERM]
chvp has quit [Quit: Bridge terminating on SIGTERM]
vaibhavsagar has quit [Quit: Bridge terminating on SIGTERM]
Ericson2314 has quit [Quit: Bridge terminating on SIGTERM]
jtojnar has joined #nixos-dev
<jtojnar>
you could want that even in the same flake – e.g. steam-run's multilib support
<domenkozar[m]>
in those cases you could do something like have two inputs with different systems
<domenkozar[m]>
there are two ways to design this: make it very generally flexible but make easy things hard, or make easy things easy and make special cases supported with extra work
<niksnut>
once upon a time I tried to enforce that committers had to use their real name (or at least it should be known)
<niksnut>
I failed
<jtojnar>
well, that commiter tried to use their real name (a mononym) and were told to switch to GitHub handle
<V>
One of these days I'll get it on my ID out of spite
<V>
At some point the number of people complaining about my name will outweigh the inconvenience of going through the legal/bureacracy involved
<eyJhb>
niksnut: I also had that discussion initially when I was added to the maintainers list. Because. I really don't want to associate my real name directly with my nick in that way :p
<eyJhb>
Waiting for you to change your real name to V V
<V>
hell no
<V>
We have a character for that: W
<eyJhb>
,pnig
<eyJhb>
,ping
<{^_^}>
pong
<V>
Actually my "real name" is randomly generated and is of zero value to anyone but the tax office and border enforcement/etc, so if people want a meaningful identity out of me they're out of luck
<V>
I literally don't have one
<eyJhb>
V: Is it?
<eyJhb>
What did you auto gerenate?
<eyJhb>
randomly generate*
<V>
a name
<eyJhb>
I would really hope you just changed it to `a name`.
<V>
lol no
<etu>
V: You're saying that your legal name kinda looks like two passwords? I have respect for that.
<eyJhb>
I don't think minimizing your online footprint can be called a myth
<sorear>
(a) it buys you *nothing*. the UMN people all used their government names and real student affiliations to submit malicious commits to linux, as was TLS extended random (b) "falsehoods programmers believe about names": what you *are* accomplishing when you try to vet people's names for being "normal enough" is structural racism and transphobia
bennofs_ has quit [Ping timeout: 248 seconds]
<eyJhb>
domenkozar[m]: (but I do get what you are saying, but I think there are more aspects to it)
<sorear>
my government name is not secret but I can't use it in most places because it *contains shell metacharacters*
<V>
sorear O'Brian
<domenkozar[m]>
sorear: how is that connected to racism/transphobia?
<domenkozar[m]>
I really don't know so I'm asking
<Synthetica>
domenkozar[m]: Things like "your name has to have at least 3 characters" leaves people called Wu out for example
<domenkozar[m]>
ohh but that's a different thing
<domenkozar[m]>
we're mixing up two things: real name and what is a valid real name
<XgF>
What is a real name? :-)
<domenkozar[m]>
I wouldn't tie that to legal documents, but whatever you ask people to refer you
<XgF>
Thats not what most people insisting people use their 'real name' mean...
<XgF>
A friend is currently having the issue that her name change document is not sufficiently official for whichever beuracrat within the German government she's currently dealing with, because name related concepts change radically every couple of hundred miles within Europe
<XgF>
And that's just Europe. Not like it gets any less complex when you involve multiple continents
<V>
that just means she has two real names, one for each country ;)
<XgF>
V: The problem is that she doesn't have a passport with the name the German government knows her by on it
<endocrimes>
XgF: I have the same problem rn and it’s very tedious
<endocrimes>
Like: “I might be deported in 6 weeks at my brexit immigration appointment” levels of tedious
<V>
"real names" are a scam invented by the state to make bureacracy easier :p
<endocrimes>
As someone who has had... enough name changes in life not entirely by choice to not even have a record of all of them anymore, I don’t disagree
pbb has joined #nixos-dev
<XgF>
endocrimes: i'm so confused, my irc client is denying your existence
<XgF>
which seems oddly topical
<eyJhb>
V: I would say that is the SSN
<eyJhb>
(that makes it easier)
<MichaelRaskin>
That would make it easier if it worked as advertised
<XgF>
What's an SSN? :-3
<V>
and if every country had one :p
<XgF>
At this point I think I have 5 tax numbers just from 2 countries
<endocrimes>
XgF: that seems very on brand for me tbh
<V>
Welcome to Britain, where there's no centralised identity database, and you aren't legally required to have a form of ID
<XgF>
V: Three of them are from the German government! The German government is worse about this :p
<XgF>
That's not including either my Health Insurance Number or my Pension Health Insurance Number
<XgF>
(I have a number for Me, I have a number for Me as a business when dealing with Germans, and a number for Me as a business when dealing with other EU member states...)
<endocrimes>
Hahaha I have to use three different names depending on which part of the DE gov I’m talking to
<MichaelRaskin>
I believe had two health insurance numbers for me as a person issued to me when moving to Germany
<endocrimes>
At least I don’t need to use my business tax ID anymore tho
<symphorien[m]>
I knew someone wih two passports with distinct names and a letter from the embassy asserting "yes yes it's the same person"
<V>
that sounds reasonable enough? legal names only really apply within their own jurisdiction
<MichaelRaskin>
As far as I understand, many parts of German state like to do bureaucracy without this liking translating to having high skills at it
<XgF>
V: Many countries require you to harmonise the name on all of your passports, which sometimes results in catch-22 situations
<XgF>
For example, if you're adual citizen with the UK, then you have to update your other passports first. I bet that at least one other country has the same rule
<puck>
XgF: there's actually a fallback in at least the UK rules; namely "not being able to change your name in the country of origin", which probably covers this
<XgF>
puck: yeah, although thats more intended for cases like you're a dual citizen with russia and they don't recognize your same sex marriage
<puck>
yeah
<sphalerite>
A Portuguese-British friend of mine just had to try to change his name in Portugal and get rejected to be able to use that rule
<gchristensen>
Regnat[m]: I'm 'round here of course if you want to talk about my feedback, is your feeling that hydra#875 is ready to go otherwise? I have mostly disregarded it as a WIP
<edef>
hmm, i did not actually realise that trick existed, or forgot it did
<gchristensen>
reponame#number?
<edef>
no, name change rejection being usable as workaround for passport harmonisation
pmy has quit [Ping timeout: 240 seconds]
<gchristensen>
oh!
pmy has joined #nixos-dev
<edef>
i have at least two names that people consider to be "real", three that the UK government considers real, one that the NL government considers real, and i can't change my NL government name to my actual name
AlwaysLivid has joined #nixos-dev
<regnat[m]>
Thanks a lot for the review gchristensen ! It's not really WIP anymore indeed (though it only provides a basic support for CA derivations and will probably require some follow-up)
<gchristensen>
I'm really keen on seeing more tests get written before merging, it would also be great to write a page in the docs on hydra + content addressed derivations
<niksnut>
ideally there's not much to say about them in the hydra docs
<regnat[m]>
Yes, that should be totally transparent on the hydra side (except for the probable quirks that will come with it, but I don't think we really want to set these in stone 😛)
<regnat[m]>
Tests are definitely worth it though
<gchristensen>
writing them down doesn't set them in stone :)
tomberek has joined #nixos-dev
justanotheruser has quit [Ping timeout: 245 seconds]
AlwaysLivid has quit [Remote host closed the connection]
AlwaysLivid has joined #nixos-dev
tomberek has quit [Quit: Connection closed]
qyliss_ has joined #nixos-dev
AlwaysLivid has quit [Remote host closed the connection]
AlwaysLivid has joined #nixos-dev
justanotheruser has joined #nixos-dev
tomberek has joined #nixos-dev
<Ericson2314>
clever: no I haven't!
orivej has joined #nixos-dev
jonringer has joined #nixos-dev
AlwaysLivid has quit [Remote host closed the connection]
AlwaysLivid has joined #nixos-dev
<clever>
Ericson2314: basically, XNU is the opensource core of the osx kernel
<clever>
Ericson2314: what if we made a nix based distro ontop of XNU, which is capable of running darwin binaries?
<Ericson2314>
clever: so you want to resurrect pure darwin?
<Ericson2314>
:)
<clever>
it doesnt need to be a fully usable distro, the main goal is "darwin" CI
<Ericson2314>
clever: that does sound great, I've thought of that before
<Ericson2314>
now what apple open sources is kinda jank, right?
<Ericson2314>
I'm thinking the second kernel will be the hardest to support
<misuzu>
clever: running macos using kvm is not an option?
<Ericson2314>
so if XNU is pretty hard, maybe tackle it 3rd
<Ericson2314>
misuzu: the impurities make for testing hell
<Ericson2314>
it's really nice to have some fully reproducible
<Ericson2314>
even if it's a bit artificial
<clever>
misuzu: then you have to deal with the entire software stack
<clever>
misuzu: and apple deciding to change things on a weekly basis
<clever>
like adding selinux style security, where sudo cant do certain things
<clever>
because a non-trusted app ran sudo
orivej has quit [Ping timeout: 240 seconds]
<gchristensen>
macos in KVM is doable but requires modules from the hackintosh community and ever hardware / software release brings up new questions of if it'll continue working
<gchristensen>
I've spent a few days this week already trying to get some new catalina machines going
tomberek has quit [Ping timeout: 240 seconds]
<domenkozar[m]>
Ericson2314: btw if you have some spare cycle it would be great to help __monty__ so that we get SDK bump before 21.05 branch off
<Ericson2314>
Which SDK?
<domenkozar[m]>
macOS SDK
<Ericson2314>
Link?
<domenkozar[m]>
__monty__ can help, don't think there's a PR yet
<Ericson2314>
i think rebasing monty's thing on top of that will lead to some goood things
<qyliss>
domenkozar[m]: in this case, my understanding is it's mostly been that macOS has been holding us back, because it was the main reason we couldn't bump the default LLVM
<Ericson2314>
already see the libunwind stuff; that's nice
<domenkozar[m]>
Ericson2314: Could we merge the macOS SDK stuff first? I'd really like to get that merged for the freeze
<Ericson2314>
domenkozar: main thing for me is both before the freeze
<domenkozar[m]>
qyliss: isn't that even more the reason so?
<Ericson2314>
I have built mine on all platforms it is ready while monty's is still WIP, but if I am mistaken about the degree of WIPness I am fine with other order
<qyliss>
domenkozar[m]: that sentence doesn't parse for me, sorry
<domenkozar[m]>
qyliss: nevermind :)
<Ericson2314>
if we do mine first I am happy to rebase monty's :)
<clever>
qyliss: thats part of the unknowns
<domenkozar[m]>
Ericson2314: I think we only have 2 days?
<gchristensen>
arianvp: it'd be neat if `nixos-rebuild boot` could fiddle set-oneshot ID Set default boot loader entry, for next boot only
<domenkozar[m]>
Ericson2314: anyway, I'll leave it up to you and __monty__ to decide how to collaborate, but note that there's a deadline so make sure you also keep jonringer in the loop
<Ericson2314>
OK
<Ericson2314>
I would merge mine right now but someone else should probably take some sort of a look
<samueldr>
#84204 boot counting, and automatic fallback, gchristensen arianvp
<gchristensen>
yeah, boot counting would also be nice
<Ericson2314>
__monty__: your PR description is a bit uncertain, but I take it we have decided on going with 11 for now?
<__monty__>
I figured that's the best way forward. The Apple Silicon work also bumps to 11. However, if getting the SDK bumped is the first priority (and CF can be fixed without bumping LLVM), we may be better off dropping that for now.
<__monty__>
When I made that decision I wasn't aiming for 21.05 tbh.
<samueldr>
gchristensen: ah, I see what you meant, "just" do a set-oneshot call, rather than changing the default
<gchristensen>
yeah
<samueldr>
ideally we'd do it through an interface that is not systemd-boot specific
<gchristensen>
indeed
<samueldr>
I mean, nixos-rebuild would end up calling bootctl with that
<samueldr>
but we shouldn't marry the end-user interface with systemd-boot
<samueldr>
IIRC grub has an equivalent too
<gchristensen>
I'm sure grub can do it
<gchristensen>
I don't know how, but I'm sure it can :)
<gchristensen>
impending potential downtime for ofborg as it gets a major update
<gchristensen>
(brief)
<Ericson2314>
__monty__: ah ok. Is there another PR or issue for the SDK? I didn't think we used the SDK
<Ericson2314>
(but I also saw some comments alluding to switching to using it or something)
cole-h_ has joined #nixos-dev
cole-h has quit [Quit: Goodbye]
rajivr has quit [Quit: Connection closed for inactivity]
<__monty__>
No PR yet no. Afaiui, it's complicated. We cobble together most of an SDK based on various versions of the open source stuff that Apple releases but the holes are filled up with the SDK.
<Ericson2314>
fun
<qyliss>
__monty__: access to path '/homeless-shelter/src/darwin-stubs/stubs/10.13' is forbidden in restricted mode
cole-h_ is now known as cole-h
<__monty__>
qyliss: That's re the darwin-stubs bump? It's referring to a local path because I was working on this alone. I'm not yet familiar with the process to get new darwin-stubs hosted and I didn't want to bother someone to update it repeatedly. You can ignore the commits after `current` for now because we'd need to get that building bootstrap-tools first.
<qyliss>
ahh, okay :)
<qyliss>
makes sense
<qyliss>
thought it was just a mistake
<gchristensen>
hexa-: here?
<hexa->
gchristensen: yep
<gchristensen>
so it looks like systemd decided to come up with new predictable names, I don't suppose networkd would handle that?
<hexa->
you can have many match types for interface configuration with networkd
<hexa->
stuff like MACAddress, PCI Path, Interface Name
<puck>
my best guess is that the kernel got updated, which kicked off a bunch of devlink stuff, which is now taken in account when generating the predictable interface names
<gchristensen>
gotcha
<gchristensen>
guarantee the kernel got updated
<puck>
the "np0" is generated from the devlink interface
<hexa->
puck++
<{^_^}>
puck's karma got increased to 44
<gchristensen>
so I assume other machines with mlx devices would also have this same change?
<gchristensen>
hrm. these other evaluators have very different configs. Let's see :)
<gchristensen>
fingers crossed, I'm doing a reboot across the fleet :)
<gchristensen>
woohoo
<primeos>
Ericson2314: Unfortunately that PR is too much for me to (properly) review atm, sorry
bennofs_ has quit [Read error: Connection reset by peer]