supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-dev
teto has quit [Ping timeout: 264 seconds]
<supersandro2000>
jonringer: How can I install pip extras in nix?
<gchristensen>
try asking in #nixos?
<samueldr>
this channel is really about coordinating the more in-depth development between maintainers, and harder questions between maintainers (but still welcome to new users!)
<samueldr>
unless I know it involves deep-rooted knowledge from few maintainers, I still go first on #nixos when asking questions
<cole-h>
If I background it, stuff shouldn't be happpening. This was more annoying when I would try to ctrl-z a build that was pegging my 4c4t CPU, but I don't mind as much now that I have 8c16t
<cole-h>
Though I understand that this isn't really possible when using the daemon
<samueldr>
the nice values in nix's conf, I think, helped in my cases
callahad87074 has quit [Ping timeout: 260 seconds]
teto has joined #nixos-dev
alp_ has joined #nixos-dev
alp_ has quit [Ping timeout: 268 seconds]
alp_ has joined #nixos-dev
jonringer has quit [Ping timeout: 264 seconds]
jonbryan has quit [Quit: WeeChat 1.9.1]
mkg20001 has quit [Quit: Idle for 30+ days]
alp_ has quit [Ping timeout: 264 seconds]
<MichaelRaskin>
cole-h: in the ideal world, what you say is actually quite possible. But quite a few changes to the daemon
<MichaelRaskin>
Daemon could track _all_ client connections waiting for some build (not only the first one, by the way)
<MichaelRaskin>
And there could be some kind of pinging
<MichaelRaskin>
If any of the connections is up and pingable, the build goes on. If all connections get broken, the build is cancelled. If there are non-broken connections, but only to stopped processes, the build is frozen.\
<MichaelRaskin>
I would surely support and test such a change, but I am not likely to be implementing it myself anytime soon
<MichaelRaskin>
Maybe you could even have a daemon proxy. This is a normal client doing realise commands for the daemon, so not risking integrity; and it accepts client requests, analyzes dependency trees and make decisions what to build in what order and when to freeze/cancel stuff
tolt____ has joined #nixos-dev
nschoe has joined #nixos-dev
tolt has joined #nixos-dev
andi- has quit [Ping timeout: 260 seconds]
justanotheruser has quit [Ping timeout: 240 seconds]
<gchristensen>
please keep a close look at the contributor named nora-puchreiner (re: https://github.com/NixOS/nixpkgs/pull/85780#issuecomment-719887350) whose github account is quite new and has only been used to contribute to nixpkgs. This is a pattern we've seen before in 0xabab and coretemp. Additionally, searching the internet for norah puchreiner brought
<gchristensen>
and notably, that is the *only* bits on the internet for that name
<eyJhb>
gchristensen: Could be the same user as coretemp? Also, I am amazed at how much they know about Nix when they start on this. Also, gchristensen is coretemp still blocked?
<gchristensen>
coretemp is still blocked
<gchristensen>
and yes it could be the same person, but I don't want to assume it is
<eyJhb>
jtojnar: I.. I don't want to make a new PR for this package :(
<eyJhb>
jtojnar: I thought that it should not end in -unstable, as then upgrade stuff? I think samueldr pointed that out
<eyJhb>
But... Intentional from nix-env behaviour??
ris has joined #nixos-dev
nschoe has quit [Ping timeout: 240 seconds]
nschoe has joined #nixos-dev
__monty__ has joined #nixos-dev
<MichaelRaskin>
Well, my github account everyone here can find easily is also only used for things related to my Nixpkgs contributions…
<gchristensen>
sure, that is different
<gchristensen>
for example, this username isn't just used on github :)
<gchristensen>
(and that isn't the only thing here)
<MichaelRaskin>
I mean, I am in favour of not reading too much into tangential details.
<gchristensen>
yea
<MichaelRaskin>
Throwing around comparisons that could be plausibly found as insulting to _both_ sides of the comparison, and pressuring a contributor towards not following a well-established and discussed to death project policy? That I also dislike…
<MichaelRaskin>
Discussion of relation to coretemp sounds like more doxing than relevant, in a sense.
<MichaelRaskin>
(I do not oppose strongly worded complaints to this policy in the many discussion threads about the policy, though — there they should be welcomed, and politely rejected)
LnL- is now known as LnL
luc65r has joined #nixos-dev
justanotheruser has joined #nixos-dev
nschoe has quit [Ping timeout: 258 seconds]
nschoe has joined #nixos-dev
nschoe has quit [Ping timeout: 240 seconds]
nschoe has joined #nixos-dev
<gchristensen>
somewhere in 733d2e9402807e54d503c3113e854bfddb3d44e0...e0ca98c recursive-nix builds through the daemon stop working
<gchristensen>
but I can't really bisect very well, as nixos-rebuild switch fails with nix-build: src/libstore/local-store.cc:483: void nix::canonicalisePathMetaData_(const Path&, uid_t, nix::InodesSeen&): Assertion `S_ISLNK(st.st_mode) || (st.st_uid == geteuid() && (mode == 0444 || mode == 0555) && st.st_mtime == mtimeStore)' failed. while checking that Nix can read nix.conf...
<Ericson2314>
can someone do the chanserv thing to make #nix-windows legit?
nschoe has joined #nixos-dev
kloenk has quit [Ping timeout: 272 seconds]
<eyJhb>
Ericson2314: Register it?
<eyJhb>
You can do that yourself
kloenk has joined #nixos-dev
<Ericson2314>
eyJhb: it says #nix-* is a namespace owned by nixos
<eyJhb>
Ohh...
<eyJhb>
Didn't know that. I am guessing srhb gchristensen can do it
<eyJhb>
I have at least observed that they have superpowers!
justanotheruser has quit [Ping timeout: 264 seconds]
jonringer has joined #nixos-dev
<V>
<gchristensen> please keep a close look at the contributor named nora-puchreiner (re: https://github.com/NixOS/nixpkgs/pull/85780#issuecomment-719887350) whose github account is quite new and has only been used to contribute to nixpkgs. This is a pattern we've seen before in 0xabab and coretemp. Additionally, searching the internet for norah puchreiner brought
<V>
okay, but I also have a github account that's very new, has almost purely been used to contribute to nixpkgs
<V>
<eyJhb> gchristensen: Could be the same user as coretemp? Also, I am amazed at how much they know about Nix when they start on this. Also, gchristensen is coretemp still blocked?
<V>
*and* I've had people surprised at how much I know about nix
cole-h has joined #nixos-dev
<eyJhb>
V: You also came and took my by storm
<eyJhb>
me by*
<V>
don't quite understand what you're saying
<supersandro2000>
I am saying one thing and then leaving this discussion: The person wrote the comment on a processor which is unfree
<V>
oh, that you're surprised by me as well?
<abathur>
V did you read the comments that were hidden?
<V>
abathur: I did
<abathur>
so I think it's less about the biographics and more about the comments
<V>
to my knowledge I haven't made unhinged comments on PR threads as of yet :)
<abathur>
but the biographics are just an extra smell
<abathur>
:]
<gchristensen>
V: did you happen to read their comment?
<abathur>
:P
<V>
gchristensen: read down :p
<gchristensen>
ok
<MichaelRaskin>
I think nobody objects that content is badly worded, and does not belong in this PR anyway. But I am not alone to dislike excessive metadata collection, so to say!
<eyJhb>
V: Yes :p
<gchristensen>
I apologize for noticing a pattern
<V>
oh, I don't take offence
<V>
I'm just really curious how many people are suspicious of me
<eyJhb>
Well, excessive requires more data I would say. Personally, I would expect someone to check further on me, if I leave comments as such
<V>
and/or confused, etc
<abathur>
V: what a suspicious thing to say :D
<V>
oh no
<MichaelRaskin>
gchristensen: noticing patterns is cool!
<MichaelRaskin>
I do not even say there is no correlation predictive power
<MichaelRaskin>
Just that I prefer to used more direct signals, especially when we are not talking about covert subversion
<MichaelRaskin>
I once did not enter display name into the mail client. This lead to a few funny conversations in the mailing list…
<abathur>
MichaelRaskin not sure I know what you mean by "more direct signals"?
<eyJhb>
`@username HEY YOU! Be nice` <-- maybe?
<MichaelRaskin>
Well, we have comments as written and we dislike them. That's signal we are actually reacting to
<MichaelRaskin>
eyJhb: are you sure you will like what the user will consider nice?
<eyJhb>
MichaelRaskin: Well, I don't take it is personal from a stranger, so I would survive
<MichaelRaskin>
You will survive anyway, so that's a low bar for deciding whether to bother!
<abathur>
I see gchristensen call people out for being rude in a pretty direct way (and appreciate it as an important part of the community's immune system!), so I doubt it's the case that he disagrees with being more direct; perhaps he didn't see the need to risk escalating the disagreement since emilazy had already called out the comment's tone?
<MichaelRaskin>
Also, it is inside a PR where there are actual on-topic things to discuss
<abathur>
nod, risks derailing it
<eyJhb>
MichaelRaskin: WELLL! Yes
<eyJhb>
Also, gravedigging?
<gchristensen>
my goal was to say: here is a new person to our community behaving badly, please watch for it and notice, and share if other people see it. the rest of it was just idle looking to see if they had a history of being a contributor elsewhere.
<cole-h>
gchristensen++
<{^_^}>
gchristensen's karma got increased to 361
rajivr has quit [Quit: Connection closed for inactivity]
nschoe has quit [Remote host closed the connection]
<worldofpeace>
gchristensen: yeah, I noticed them on that pr yesterday
<MichaelRaskin>
gchristensen++
<{^_^}>
gchristensen's karma got increased to 363
<worldofpeace>
overall gchristensen choice of calling out people with patterns of bad actors is something I appreciate greatly and hope we can all do collectively. because if it's just the person who's been affected by their words trying to do it... it's not enough. And absolutely in cases where escalating the situation is not really needed, just being aware of those individuals, and to watch out for yourself and everyone else. I
<worldofpeace>
also do this and inform people who could likely become targets of certain people. Also, if you've actually been personally attacked by people on the internet... you will quickly see this is essential to have a healthy community.
<V>
I wasn't complaining, FWIW
<worldofpeace>
never should we tolerate people who's words are always less than kind, less than helpful. if will spread and someone will have an experience with that person and get the wrong idea about nix
<worldofpeace>
V: lol, I just felt like saying this here FTR
<worldofpeace>
✨ gchristensen
<{^_^}>
gchristensen's karma got increased to 364
jonringer has quit [Ping timeout: 264 seconds]
justanotheruser has joined #nixos-dev
* gchristensen
git bisects while he git bisects to avoid `git skip`ing around a bug 50 times in 54 commits for
<cole-h>
x(
<samueldr>
as I wished for a while ago: multi-status git bisect!
<samueldr>
what if instead of good/bad, [or equivalents] we would be able to say --terms=works,broken,unexpectedly_broken
<gchristensen>
skip-for-bug-1, bad-for-bug-2
<samueldr>
or uh, works,unexpected,broken; as it'd (hopefully) work as last that works, unexpected span, and first broken
<gchristensen>
also I'd like to be able to trisect
<MichaelRaskin>
In what sense?
<gchristensen>
my computer is fast enough to build and run my test, two at a time
<MichaelRaskin>
I think last time I considered git bisect, I tried, then ended up just manually keeping detailed records
<gchristensen>
so I'd rather make progress faster
<MichaelRaskin>
Hmm indeed, parallelism is easier than sequential performance
<gchristensen>
or maybe I have a big ol cluster of machines and I'd like to just get an answer nowish, so git n-sect --n=200 and divide my 1000 commit range in to 200 chunks
<samueldr>
thinking a bit more: adding terms live is probably helpful
<samueldr>
"oh, it fails it with abcdefg again, git bisect --add-term fail-abcdefg"
<MichaelRaskin>
If you have 200 machines in a single cluster, the company should be able to hire a recent graduate to write this proper bug-locator DVCS script already
<gchristensen>
I can have 200 machines for an hour but can't afford a grad student
<MichaelRaskin>
Well, if you can have 200 machines for an hour, you are not the one who has 200 machines in a cluster
<gchristensen>
yeah but Amazon isn't going to give me a grad student :P
<andi->
Just get a grad student to hack into 200 machines?
<andi->
lease them out and use for an hour
<gchristensen>
but you can get 200 pretty okay machines for 1h, all at once, and have it cost $25-$50
<sphalerite>
gchristensen: are you sure? Maybe Amazon Mechanical Turk has an option for that.
<gchristensen>
they'll only give me a grad student for about 30 seconds at a time :(
<sphalerite>
oh ok
<sphalerite>
110348:2018-02-18 19:00:26 Peng_ Cloud quantum bogo bisect: Use cloud computing to test every commit at once. :P
<sphalerite>
it's evergreen.
<gchristensen>
like the lambda edition of grad student assigrments
<sphalerite>
(this was in the context of me asking about n-section back then :D )
<gchristensen>
:D
<MichaelRaskin>
Well, if Amazon has not yet written a bisection tool, I guess there are no drawbacks in anti-trust splitting them into enough 200-servers-owning companies that at least one ends up writing such…
<gchristensen>
haha, yes
<MichaelRaskin>
Now that I think of it, it is not implausible that I get ~200h of BSc student work in the near future. And if they turn out to be practical-minded (technically speaking, we are the chair of _theoretical_ CS, but…) and nobody at the chair is eager to sketch me a thing to implement…
<supersandro2000>
little OT question: school student or university student?
<supersandro2000>
(translates not very well for me)
<gchristensen>
university
<MichaelRaskin>
University student
<MichaelRaskin>
End of third year
<supersandro2000>
When people only want to university students I always feel a bit underestimated as apprentice
<supersandro2000>
*a
<MichaelRaskin>
I guess back in Moscow I could try to get ~ 100h of high school student work on that…
<MichaelRaskin>
(and this would have a pretty high chance of success at least up to some milestones)
<MichaelRaskin>
(I mean, 100 h from zero is not _that_ much time to make a thing work reliably!)
<supersandro2000>
Zero in CS or in the problem?
<MichaelRaskin>
In the problem
<MichaelRaskin>
Both university and highschool students implied are the ones with experience debugging stuff like graph algorithms
<supersandro2000>
that sounds reasonable
sphalerite is now known as spookyscarysphal
cole-h has quit [Ping timeout: 264 seconds]
scott has quit [Quit: bye]
scott has joined #nixos-dev
v0|d has joined #nixos-dev
<eyJhb>
How did this go from git bisect, to enslaving grad students?
<worldofpeace>
eyJhb: yeah, that has me feelin some type of way
* V
raises an eyebrow
<gchristensen>
I don't think that it as *at all* what happened there