<siraben>
The review process for `stdenv.lib` → `lib` PRs relied a lot on ofborg's evaluation, which was found to be incomplete in certain cases, thus the solution is to ensure more complete evaluation to reduce breaking PRs.
<adisbladis>
Another thing we need, especially for projects like poetry2nix is a way to run the exact same checks externally
<adisbladis>
Otherwise it's really annoying to try to get something ready for a nixpkgs PR
<siraben>
FWIW `stdenv.lib` will still exist for the next release and will be removed in 21.11, so there shouldn't be external breakage due to this change.
<rnhmjoj>
anyone here using thinkfan? i have a somewhat large PR that needs some feedback: #111624
<supersandro2000>
if you eval nixpkgs without aliases and checkMeta you will catch most errors
<supersandro2000>
siraben: but if you upstream something ofborg will most of the time complain
<supersandro2000>
I thought it should also complain in the poetry update
<symphorien[m]>
<rnhmjoj "anyone here using thinkfan? i ha"> I do but i'm short on time. Maybe this weekend.
<philipp[m]1>
I'm reactivating an old host and it seems like the gitlab-runner module has been changed significantly from 20.03 to 20.09 (with breakage of previously usable features) and there is neither a mention in the release notes, nor any migration path given.
<{^_^}>
#108763 (by misuzu, 3 weeks ago, merged): nixos/gitlab-runner: add support for 'docker+machine' and 'docker-ssh+machine' executors
<philipp[m]1>
I don't need any help, I just wanted to note this here so maybe the release notes can be complemented.
AlwaysLivid has joined #nixos-dev
AlwaysLivid has quit [Ping timeout: 268 seconds]
orivej has joined #nixos-dev
jonringer has joined #nixos-dev
<gchristensen>
it looks like GitHub might have fixed the etag issue?
<immae>
gchristensen: A patch made it to nixpkgs recently, so if you upgraded recently it might be that rather than github :)
<gchristensen>
I haven't upgraded though :)
<immae>
(a patch for nix)
<immae>
Then it’s probably github :p
mkaito has joined #nixos-dev
mkaito has joined #nixos-dev
mkaito has quit [Changing host]
<infinisil>
gchristensen: Oh I got the DBIx error. I don't think this needs to be fixed, it's probably just a transient one, but for reference: https://paste.infinisil.com/gHc5zvBX3M
<gchristensen>
huh
<infinisil>
This appears in the evaluation errors tab of a jobset, hydra version 1d45b63
<abathur>
do we know anyone using Nix + macOS in EC2?
<abathur>
and secondarily, in VMs in general; we had someone turn up in nixos and nix-darwin reporting some trouble installing Nix on EC2 that echoed some trouble sandro had a few months ago installing it in a VM
<andi->
gchristensen: I hate to bring it to you but I think hydra can't rename jobsets anymore. I just renamed one and the UI is showing one name and the evaluator is using another one
<gchristensen>
hmm
<gchristensen>
beyond confusing output, I'm not sure that is actually a significant problem since it is using jobset IDs everywhere of significance ... and I sort of expect that in prior versions it would just crash and restart to "deal" with it... what is your experience?
<gchristensen>
I'm guessing it'd crash because it would try to read or write with the name and get an invalid reference
<gchristensen>
andi-: so I guess I'm wondering if you can confirm the problem is scoped to the messages it is printing, or if the records recorded are tainted
cole-h has joined #nixos-dev
<ashkitten>
niksnut: what was the original intent behind removing the ld.so.cache path from glibc?
<gchristensen>
it is hard for it to be correct, since it is inherently an impurity
<clever>
ashkitten: each binary would basically need its own cache
<ashkitten>
ah right, i see
<ashkitten>
i think what i want to do is, make it look for `ld.so.cache` but not generate its own
<ashkitten>
that way we can put an `ld.so.cache` inside a fhsenv (steam) and it will find that
<immae>
gchristensen: about https://github.com/NixOS/hydra/pull/856 I pushed a commit where I delete + recreate the constraint on schedulingShares. This because there is no syntax to say "add constraint if not exists" (but there is a "drop if exists"), and checking for its existence would make the query much harder to read. Is it okay for you that way, or do you prefer the more complex one?
<{^_^}>
hydra#856 (by immae, 2 days ago, open): Fix check in jobsets
<gchristensen>
immae: hmm... right. I think we should stay simple, though it strikes me picking a new name lets us create the new one and drop the old one and not having a point where we're not enforcing the check
<gchristensen>
this feels paranoid to me, but I think it is good to just not tempt fate, you know?
<immae>
I know, and I can understand that :p
<immae>
Do you want me to write the complex version and then decide based on how you feel about it?
<gchristensen>
no, I'd rather go the cheap and cheerful way of having a different name -- like jobsets_schedulingshares_nonzero
<immae>
(when I say complex it’s because there will be a DO...IF...THEN...)
<immae>
ah yes
<immae>
giving a name that we’re sure wouldn’t exist
<gchristensen>
yeah, exactly
<immae>
I would add a "check" at the end, since constraint names are global, "jobsets_schedulingshares_nonzero_check"
<gchristensen>
ah, yeah, good idea
<immae>
I pushed the change
<immae>
I tested it on a brand new database and on a database at previous step + last migration
orivej has quit [Ping timeout: 240 seconds]
<immae>
You said you had a way to test it on a copy of current prod?
<immae>
(I could mimic its state myself but it’s even better if you have a true copy)
<gchristensen>
great! yeah, I have a literal copy of it here, I'm just finishing up another patch, and I'll take a look at that PR
<gchristensen>
thank you so much, I really appreciate it :D
<gchristensen>
I'm glad you could help figure this out with me
<immae>
You’re welcome :)
<gchristensen>
just learned about Test::More, seems nice compared to Test::Simple
<gchristensen>
I've learned more Perl than I really wanted but it is what it is
<supersandro2000>
abathur: I think it was something like /nix/var/nix/profiles/system/sw/bin/sh
<supersandro2000>
but it only ever happened when running in daemon mode
<gchristensen>
immae: thinking through these constraints is stressing my brain :P
<immae>
gchristensen: ah, do you need help to get a grasp of it?
<gchristensen>
I want to try to get myself there first
<immae>
Sure :)
<gchristensen>
I just need to write it down
<immae>
I’m around if you need
<gchristensen>
thanks!
<gchristensen>
the migration is pretty reasonable and easy to understand, though, pretty good!
<ris>
fetchpatch silently stripping binary sections is a particularly unhelpful feature imho
<immae>
cool that’s a good point
tilpner_ has joined #nixos-dev
<cole-h>
Just FWIW: redeploying ofborg. PR statuses have the potential to be delayed, but I don't foresee anything going wrong.
tilpner has quit [Ping timeout: 265 seconds]
tilpner_ is now known as tilpner
BaughnLogBot has quit [Ping timeout: 246 seconds]
pmy_ has quit [Ping timeout: 256 seconds]
pmy_ has joined #nixos-dev
<gchristensen>
immae: I really don't like to have so much discussion on what is a small PR. I don't mean to pick on you, but understand it for myself too. I hope you're not annoyed by it
<immae>
gchristensen: no it’s not a problem
<immae>
Code clarity is important, and as a mathematician some things like the one I wrote there might seem natural to me while they’re not :)
<gchristensen>
ah! yeah, I don't often face very complicated rules like that, I'm definitely out of practice
<immae>
Nice script by the way
<gchristensen>
thanks :) I started doing it in Excel and then realized I could just try them all against a real database ^.^
<immae>
ha ha
<immae>
I’m not sure about how to understand "the `known` check was just to play further". Does it mean that I should not add the check (and thus allow type = 2 + strange combination of flake/nixexprInput/nixExprPath) because anyway type will always be 0 or 1?
<gchristensen>
ah ... yeah, let's add that check too, it'll be g ood
<immae>
ok
<immae>
(note that my version didn’t care of what happened when type was not 0 or 1, because it was unspecified in the stated constraints :p )
<gchristensen>
hehe, yeah ... as you correctly noted, the constraints are too generous to start with :P
<immae>
gchristensen: I updated the PR and tested again on the two cases I was able to test ;)
<gchristensen>
nice
<immae>
(and I confirm mathematically that the checks do honor the expressed constraints)
<gchristensen>
nice! how did you do that?
<immae>
basically the same as what you started to do in excel, but with a bit of practice of that kind of situation ;)
<gchristensen>
ah, nice
<immae>
a table where you make every possible list of [type, nixExprInput is null, nixExprPath is null, flake is null, calculated_check, wanted_result] and compare the last two items
<immae>
(and discard non-relevant values, like type > 1)
<cole-h>
Update: ofborg deploy finished. Nothing out of the ordinary.
<gchristensen>
I should probably rewrite that python script in perl and add it to the test suite in case a third type appears
<cole-h>
RiiP
<cole-h>
(both meanings)
<cole-h>
immae++ gchristensen++
<{^_^}>
( *∀*)y─┛ 420 blaze it
<{^_^}>
gchristensen's karma ^^
<{^_^}>
immae's karma got increased to 28
<cole-h>
lmaoooooooo
<gchristensen>
omg haha
<cole-h>
We're all professionals here :DDD
<immae>
:D
BaughnLogBot has joined #nixos-dev
<immae>
gchristensen: if a third type appears it might become much more difficult to write new checks :p
<immae>
(hmm hold on)
<immae>
Ah no your version is quite easy to extend indeed
<immae>
(well not sure, let’s wait until a third type appears :D )
<gchristensen>
they'll just get ... long.
<gchristensen>
I think
<immae>
yes I think you’re right we’d just need to add the condition on the third type at the end of each line and add the type = 3 line. (I somehow expected that it would weaken the constraint and allow unwanted combinations, but I’m probably wrong here)
<immae>
Your script would be more useful than my brain if we had a third type :p
<gchristensen>
:D
<infinisil>
{^_^}: NICE
mkaito has quit [Quit: WeeChat 3.0]
ashkitten has joined #nixos-dev
worldofpeace has joined #nixos-dev
worldofpeace has quit [Changing host]
Irenes[m] has joined #nixos-dev
Irenes[m] has quit [Changing host]
BaughnLogBot has quit [Ping timeout: 276 seconds]
BaughnLogBot_ has joined #nixos-dev
BaughnLogBot_ is now known as BaughnLogBot
jonringer has quit [Remote host closed the connection]