<garbas>
worldofpeace: i was hoping to have a new downloads page. but i wouldnt like to block release because of it.
<garbas>
worldofpeace: but, it would be great to bring marketing team to help release team with announcement and highlights for 20.09.
<gchristensen>
+1
<garbas>
worldofpeace: will there be any postmostem meeting for the release?
<garbas>
usually it is good to have postmortem as soon as possible to keep the ideas fresh. but i'm not sure how would everybody react on yet another meeting.
<worldofpeace>
We were going to have another go/no-go on wednesday. that's in 5 days. Not sure about postmortem
<worldofpeace>
(I find postmortem a weird word because it's like after death, even if used as after)
<worldofpeace>
garbas: though I think more regular things are actually more important than the big ones. those are only productive after
<garbas>
postmostem, i wonder why we use such terms :) i like the story for "canary deployments"
<garbas>
a more regular meeting would be even better
<garbas>
meetings*
<abathur>
"process review"? :)
<flokli>
garbas: we can do a 20.03 postmortem once 20.09 has been released and is a month out ;-)
<samueldr>
flokli: I understand that you're volounteering for releasing 20.09?
<samueldr>
;)
<gchristensen>
w00t
<samueldr>
(btw I'm not being honest here)
<flokli>
uff, no. Already spending wayy too much of my evenings on nixpkgs stuff
<flokli>
I need to get my todo count to 0
<samueldr>
that number can decrease?
<gchristensen>
last I popped 1 off 3 went on
<flokli>
I managed to get from Inbox 300 to Inbox 100 while staying at home.
<samueldr>
that doesn't sound healthy
<flokli>
I wouldn't consider "amount of unactioned mails in mailbox" a proper metric for todos
<flokli>
But it's one part of the equation ;-)
<gchristensen>
12368...
<flokli>
gchristensen: that's your email count?
<gchristensen>
yea
<flokli>
unread?
<samueldr>
1196 unread, but anything github is sent to a label
<samueldr>
2157 :o
<flokli>
I was just gonna ask, did you enable ALL the github notifications?
<flokli>
:-D
<gchristensen>
unread
<flokli>
dude
<gchristensen>
want some/
<flokli>
nah, you can keep them
<gchristensen>
okay, I'm good at that
<flokli>
:-)
<gchristensen>
they lose their value if you open them. mint condition!
<flokli>
I can "unread" them again
nh2 has quit [Ping timeout: 246 seconds]
nh2 has joined #nixos-dev
elvishjerricco has quit [Ping timeout: 272 seconds]
elvishjerricco has joined #nixos-dev
<jtojnar>
ugh, planned to be here for the meeting but encountered some snow when cycling, which made the trip two hours longer
<gchristensen>
ouch!
<cole-h>
Suuuure, that's what they all say... :P
<cole-h>
At least you got back safely (or so I assume)
<jtojnar>
yeah, it was cool, trying to ride few kms in the snow, just tiring, skidding all the time
<jtojnar>
worldofpeace: I am not aware of any critical GNOME issues
<abathur>
gchristensen floki samueldr: I used to give an inbox-zero student in my grad program anxiety whenever I left my gmail tab open for a bit during a presentation or meeting :]
<abathur>
at the time it was only like 14k unopened
<abathur>
the total used to be more obvious before the inbox/social/promotions/updates/forums split
<abathur>
I guess I've got 41618 unopened messages across all 5 bins, now?
<abathur>
every few years I'll go dump several thousand messages with a search for facebook/twitter/groupon/etc.
<abathur>
though generally gmail storage growth seems to outpace my accumulation rate
<aanderse>
ps. gchristensen is something going on with bot? i pinged you on a PR
<gchristensen>
link?
<gchristensen>
github has been breaking their API
<gchristensen>
if anyone sees an ofborg-internal-error, please ping me as quickly as possible -- I can't keep many logs due to how much logging happens, but it usually indicates breakage in the github api
<gchristensen>
I don't suppose anyone has easy recommendations on setting up centralized logging for nixos in like, a few hours
<gchristensen>
without a ton of components
<hexa->
thinking loki is the smallest scalable solution
<hexa->
that might be fun to use :)
<hexa->
depends on what you expect from such a service
FRidh has joined #nixos-dev
<gchristensen>
I'd like to be able to ideally some primitive amount of search over 12 hours of data across like 10 machines. push or pull is fine. integration with journald would be ideal
teto has joined #nixos-dev
<gchristensen>
hexa-: do you have a favorite quick-start or reading material for loki to start with?
<hexa->
hmm
<hexa->
lemme check
<hexa->
i have some config lying around for that
<hexa->
with syslog forwarding
<MichaelRaskin>
Finally I get this quasi-GitHub fishing too
<gchristensen>
MichaelRaskin: congratulations!
<gchristensen>
hexa-: anything you could share would be great :)
<MichaelRaskin>
(also, HTML-only from GitHub? Really? also, what the hell with the COVID text? I know, I know, trying to throw spamfilters off the trail — and should be fine print, but I view in terminal…)
<hexa->
we tried collecting openwrt router logs with that
<gchristensen>
hexa-: so rsyslogd is forwarding to promtail, promtail is slurping up /var/log and journald and syslog on 1514, and forwarding to loki?
<hexa->
gchristensen: yep
teto has quit [Ping timeout: 272 seconds]
<hexa->
very interested in the systemd journal compat part
<hexa->
haven't looked into that yet
<gchristensen>
cool
<gchristensen>
thanks! this looks pretty easy!
teto has joined #nixos-dev
<qyliss>
HTML-only is usually my first sign that something is wrong too
<qyliss>
make it stick out like a sore thumb
<gchristensen>
also, seriously, if you can, get a U2F key and use it instead of TOTP when possible
<hexa->
gchristensen: how do you backup your u2f key?
<gchristensen>
I don't, I have the emergency tokens saved in pass
<hexa->
obviously a trick question :p
<gchristensen>
:)
<MichaelRaskin>
gchristensen: I think it is _more_ useful to automate password manager actually using the URL domain from the browser window
<gchristensen>
yes that helps, but U2F doesn't depend on users doing that, and doesn't fall for the "oh my password manager is buggy" problem
<MichaelRaskin>
Because that also breaks phishing on the websites not supporting all the buzzword bingo
<gchristensen>
U2F isn't only buzzword-security
<gchristensen>
it is seriously effective
<gchristensen>
for example, it has no shared secret, and is bound to the domain origin. you get the domain origin but you have the shared secret. TOTP has a shared secret too
<MichaelRaskin>
At that point I want to observe that telling a browser to authenticate with U2F permits the browser to authenticate with U2F to anything (which is pretty likely to be exploitable in terms of iframe-based user identification). Is there a software U2F implementation for Linux with proper filtering?
<sphalerite>
o/ people of #nixos-dev but particularly niksnut: I'm thinking of submitting an NGI Zero project to improve nix's distributed build support. My main question is if anyone's interested and would support upstreaming an effort along these lines or if anyone has any further input :)
<gchristensen>
sphalerite: niksnut is only highlighted if your message begins with "niksnut: "
<gchristensen>
U2F doesn't work requires user interaction and doesn't work with cross-domain iframes
<MichaelRaskin>
Requires — as implemented by browsers? I mean, by now Web is too much of a garbage pile to believe in proper implementations on any side.
<gchristensen>
no, implemented in the hardware token
<MichaelRaskin>
Ah, that's better
<MichaelRaskin>
But I won't bother with HW tokens unless employed by something that already employs people making sure access tokens work perfectly.
<MichaelRaskin>
gchristensen: no, unless _I_ am employed by entity already paying people to make sure that tokens work ideally
<gchristensen>
ah :)
<MichaelRaskin>
Google using something is a complicated signal from my point of view, actually
<MichaelRaskin>
Not purely positive
<gchristensen>
hehe, yeah same
<gchristensen>
"Facebook uses btrfs" good signal: they use it, less good signal: at massive scale you can afford more failures in the face of certain benefits
<MichaelRaskin>
True signal: your HDDs will fail, sooooo when did you manually look up the status of backup cronjob?
<MichaelRaskin>
Also, Google is just a huge bunch of «doing things not _quite_ correctly»
<sphalerite>
niksnut: (and others) I'm thinking of submitting an NGI Zero project to improve nix's distributed build support. My main question is if anyone's interested and would support upstreaming an effort along these lines or if anyone has any further input :)
<sphalerite>
gchristensen: thanks!
<tilpner>
sphalerite: Perhaps it would make sense to name a few specific things you would like to improve? (Or maybe it's just me who has no idea what you mean specifically)
<MichaelRaskin>
I also have some stuff (which is maybe not so much about distrivuted builds as about general scheduling), not written down mostly (but could sketch down if useful)
<MichaelRaskin>
And happy to read for consistency/orthogonality
johnny101m has joined #nixos-dev
teto has quit [Ping timeout: 260 seconds]
teto has joined #nixos-dev
<Profpatsch>
sphalerite: improving the build scheduler?
teto has quit [Ping timeout: 272 seconds]
cole-h has joined #nixos-dev
justanotheruser has quit [Ping timeout: 260 seconds]
<sphalerite>
tilpner: a number of things around how remote builds are performed, to improve both performance and usability
<sphalerite>
gchristensen: yes, I have those on my list of things to consider already because I was very intrigued the last time I saw you sharing it :D
<sphalerite>
gchristensen: thanks for those!
<sphalerite>
Profpatsch: the build scheduler and its configuration
<colemickens>
!!
<colemickens>
I have been writing my own scripts wrapping nix and nix-build in hte last few days dealing with remote builds.
<colemickens>
I have /lots/ of thoughts.
<colemickens>
those notion's should be linked from any nix distributed builds wiki pages :/
<gchristensen>
oh?
<colemickens>
just skimming through it is showing me stuff I've painfully learned over the past few days about how remote builds behave, differences in ssh/ssh-ng, substitution, build option propagation
<gchristensen>
it didn't occur to me I was writing documentation
<samueldr>
colemickens: you probably should do it then :)
<Profpatsch>
sphalerite: great, I’d love to see improvements to the scheduler. Especially removing the property that it has to copy all intermediate store paths to the host machine after every step would be great
lovesegfault has joined #nixos-dev
<lovesegfault>
The size of the hydra queue is giving me anxiety
<samueldr>
ah, easily explanable, some user-specific evals
<samueldr>
or use-specific
<samueldr>
IIRC they have a lower share allocation, so it affects the main things less
<samueldr>
and 36987 is on darwin, which would only hold up nixpkgs-unstable, nixos-unstable won't be held up by that
<lovesegfault>
What's this structured-attrs jobset?
<samueldr>
evaluating what I understand is an upcoming refactor of the way attrsets are given in derivations
<lovesegfault>
b/c of some networkmanager-fortissvpn pkg
<lovesegfault>
I don't even use that, wonder where it's coming from
<cole-h>
globin: So with structured attrs as they are now, are env vars supposed to go in `env = { ENV1 = "..."; ENV2 = "..."; };`? Just looking through your recent "fix" commit, it looks like that's the case
<lovesegfault>
Eh wtaf? enabling the NeworkManager module brings in a bunch of garbage in a non-overridable way
<mjsir911>
Hello all! Can I do anything to help with getting 72296 reviewed?
<lovesegfault>
which made it impossible to have networkmanager without all those addons
<lovesegfault>
which in turn makes it impossible to use nm in a system with noXlibs = true
<worldofpeace>
lovesegfault: I wasn't considering that. It was a bad issue because you couldn't evaluate those desktops together, and I wasn't considering testing that because there's no whispers of it in the module. We could add it back, with the suggestions infinisil made so it actually works
<lovesegfault>
worldofpeace: I'll try and get that done this weekend if I have some time :)
<lovesegfault>
Thanks for helping trace the reasoning :D
<worldofpeace>
lovesegfault: no problem. I'm surprised I remember in such detail :)
<infinisil>
gchristensen: If you have time it would be nice to get your approval for #82461
<{^_^}>
#83618 (by FRidh, 1 week ago, open): Staging next
justanotheruser has quit [Ping timeout: 256 seconds]
<gchristensen>
these are very annoying :( I cannot believe github has broken their api so much. jtojnar: if they are more than 5-6h old, simply remove them and re-evaluate. I'm working on collecting the logs better so I Can go further back
<jtojnar>
thanks
<domenkozar[m]>
vscode is going to ship extension api in next version that allows extensions to modify shell environment and propagate that to all the editor bits