supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-dev
jonringer has quit [Remote host closed the connection]
ris has quit [Ping timeout: 246 seconds]
jonringer has joined #nixos-dev
rajivr has joined #nixos-dev
<cole-h>
Has it ever been suggested to add a `default.nix` to all the various "sections" of Nixpkgs (e.g. `tools/default.nix`) that imports that section, to keep the `top-level/all-packages.nix` somewhat organized?
<samueldr>
I seem to recall that yes, but I don't remember what the outcome of the discussion was
<cole-h>
Then (with the help of some nix parser), it would be easy to sort all the attrs however we want (probably alphabetically)
<cole-h>
mostly wondering because I always see things getting added to the end of the file, even though "It is sorted by categories corresponding to the folder names"
<cole-h>
would be nice to return to that
<siraben>
"error: invalid value 'c++20' in '-std=c++20'", what stdenv do I use for this?
<siraben>
Yeah it'd be nice to split up all-packages.nix
<rmcgibbo[m]>
FYI: I've been working on some new code that will slightly change the behavior of the r-rmcgibbo bot. Instead of building _all_ attrs whose store paths are changed by a PR, it will restrict its attention to the attrs that were (a) directly modified by the code in the PR (b) direct dependencies of (a), and (c) direct dependencies of (b).
<rmcgibbo[m]>
This will be coupled with some new code that further filters the set of (a) + (b) + (c) to identify a subset that I believe I can finish building in <2 hours. Skipping packages that are more than 1 full dependency removed from the changed expressions and that are estimated to take too long for me to finish without a timeout will allow the bot to cover a much higher percentage of PRs that target master, possibly all.
<rmcgibbo[m]>
siraben: that's weird, no? isn't the normal stdenv on gcc-10 which has most of c++20 support?
<rmcgibbo[m]>
oh that's beyond me then. a new linker too... way past my paygrade :P
Taneb has quit [Quit: I seem to have stopped.]
Taneb has joined #nixos-dev
<siraben>
Heh, no worries.
<siraben>
rmcgibbo: I appreciate your bot by the way!
<rmcgibbo[m]>
thanks! i hope it's useful.
<siraben>
Are there potentially security problems by just running the builds blindly?
<rmcgibbo[m]>
lol yeah definitely.
<siraben>
But network access is disabled right?
<siraben>
In the sandbox
<rmcgibbo[m]>
but, you know... yolo?
<siraben>
hehe
<rmcgibbo[m]>
right, the nix sandbox helps.
<rmcgibbo[m]>
i run it fresh amazon vms and they don't really have any secrets on the box. after the build finishes the vm gets destroyed.
<siraben>
Ah great.
<siraben>
I wish I knew enough devops knowledge to orchestrate something like that
<rmcgibbo[m]>
eh, i'm making it up as i go along. my real background is computational chemistry.
<abathur>
rmcgibbo[m]: at least your background has computational in the name :)
<abathur>
I guess there aren't any strictly devops degrees anyways
<abathur>
or at least, it would surprise me a bit
orivej has quit [Ping timeout: 245 seconds]
<samueldr>
sshhh, they'll hear you
<abathur>
I do kinda want to see educated devops grads
<abathur>
and I can't quite decide if this is because I suspect the academy would have better ops, or that *ops just feels like something that needs to emerge from practice
<abathur>
er, not phrased clearly
<samueldr>
locally, "higher education" for compsci seems to end up lowering the quality of the many individuals :/
<samueldr>
from anecdotal evidence
<abathur>
I wouldn't really know, but I get that impression
<siraben>
"seems to end up lowering the quality of the many individuals", what do you mean?
<samueldr>
and some peeps I know who started "higher education" in compsci didn't depict a good image of the program
<abathur>
I started in comp-sci, but my undergrad and masters degrees are both in poetry :)
<siraben>
maybe → #nixos-chat
<samueldr>
oops
<abathur>
true
<samueldr>
off-by-one
* abathur
throws down a smoke-bomb and pretends this never happened...
<Rovanion>
They can't find /etc/slurm.conf and therefore don't work when run in the NixOS testing infrastructure.
jonringer has quit [Ping timeout: 264 seconds]
evils has joined #nixos-dev
<siraben>
Is there a way to grep for a specific string "Failed: 0" in check phase?
<siraben>
The test script for a package doesn't exit with 1 if one or more tests fails
<Rovanion>
siraben: Just blind guessing here: The success() call returns stdout of the subprocess, I believe. So you you should be able to write `if "Failed: 0" in success("make check"): abort_abort()`.
<Rovanion>
siraben: Wait, sorry. I was thinking about the NixOS test infrastructure where the tests scripts are written in Python. Disregard the above.
<gchristensen>
I wonder: if we enabled emails from hydra *now*, would it be a big surprise to so many users as to be perceived as generally unwanted?
<gchristensen>
if we'd never disabled emails it would have been a "this is just the way it is" kind of thing, with a lower chance for collective unhappiness to move the needle
<immae>
Whe you say "never disabled e-mail", do you mean that hydra used to send e-mail, and it is currently broken?
<gchristensen>
yeah
<gchristensen>
well, disabled
<gchristensen>
it would send email when things broke
<immae>
ah
<gchristensen>
at the size of nixpkgs now, we've probably grow enough to require proper opt-out / unsubscribe stuff, where before it was such a small thing
<immae>
If it didn’t send e-mail because broken and you fixed it, then I would say "it’s how it was meant to be, it was just broken". But if it was purposedly disabled I don’t know, I’d say it’s a change of "contract" no?
<gchristensen>
well ... it was disabled because of a mistake in a release which caused a *lot* of unnecessary emails to go out, so ... :)
<gchristensen>
anyway, people wish it'd come back in some form
<immae>
Then go ahead and say sorry if too much people complain :)
<ekleog>
I think the “hydra sending mail” ship has sailed now (still personally wish for it to come back but there are reasonable objections to it); something we could hope for would be eg. RSS feeds of breakage per-package and then you can subscribe to it yourself
<gchristensen>
sounds expensive for hydra to generate and serve the polling for, what about some sort of message queue thing
<ryantm>
supersandro2000 right. R-ryantm only PRs updates that succeed.
<symphorien[m]>
rss has headers to tell clients to only poll, say, once a day or anything you want, really
<NinjaTrappeur>
A rss/atom feed is also resonably easy to cache.
<NinjaTrappeur>
+1 for the hydra notifications. Several of my tests went red for a long time un-noticed.
<NinjaTrappeur>
(by easy to cache I meant not necessarily have to be served by the hydra instance itself)
<{^_^}>
hydra#886 (by grahamc, 37 seconds ago, open): statsd: add a chance to set hostname and port in hydra.conf
rj has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
rj has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
rj has joined #nixos-dev
mkaito has quit [Quit: WeeChat 3.0.1]
rj has quit [Remote host closed the connection]
rj has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
rj has joined #nixos-dev
<supersandro2000>
gchristensen: generate the feeds once and then cache them
<gchristensen>
it is not so trivia0l
<supersandro2000>
if we generate them for every GET hydra will go down
<supersandro2000>
not as trivial as generate the feed with crontab and serve it with nginx?
<supersandro2000>
also I would make mails opt in until they are proven to work
<supersandro2000>
and add a max amount that is send out per day
<supersandro2000>
or maybe send them in batches aka that broke today
cole-h has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
rj has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
mkaito has joined #nixos-dev
mkaito has joined #nixos-dev
rajivr has quit [Quit: Connection closed for inactivity]
stigo has joined #nixos-dev
orivej has quit [Ping timeout: 245 seconds]
rj has joined #nixos-dev
justanotheruser has quit [Ping timeout: 264 seconds]
justanotheruser has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
rj has joined #nixos-dev
ris has joined #nixos-dev
johanot has quit []
johanot has joined #nixos-dev
drakonis has quit [Remote host closed the connection]
drakonis has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
rj has joined #nixos-dev
catern has quit [Remote host closed the connection]
terrorjack has quit [Ping timeout: 246 seconds]
terrorjack has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
rj has joined #nixos-dev
maxine has joined #nixos-dev
rj has quit [Ping timeout: 268 seconds]
rj has joined #nixos-dev
<samueldr>
is plasma5 with wayland something that's expected to work with NixOS?
<samueldr>
I see things seem to be packaged for it, but testing in a vm seems inconclusive... but I might be starting it wrong
<samueldr>
using startplasma-wayland
<samueldr>
though a simple nix-shell, kwin seems to be able to start a wayland compositor on x11, but the packaging of plasma AFAIUI does not allow using it through a nix-shell
<adisbladis>
samueldr: I had it working at some point