supersandro2000 has quit [Remote host closed the connection]
supersandro2000 has joined #nixos-dev
Synthetica has joined #nixos-dev
<eyJhb>
Updating LaTeX and adding a new nodePackage... Might as well put a recompile of coreutils in while I am at it
WORLDofPEACE has left #nixos-dev ["User left"]
<supersandro2000>
latex just has 1000 packages that are 5kb big
<eyJhb>
supersandro2000: Currently updating it, so have to get every single annoying package. And that somehow takes forever. Not sure what it does when `building` them however..
<eyJhb>
Ie. I have to recalculate each hash of everything. So a normal build would actually be much quicker.
<primeos>
Synthetica: Yes, we can likely merge it later today (Sway's release time was a bit unfortunate for me but I should have more time in approx. 5h)
<eyJhb>
supersandro2000: still running
justanotheruser has quit [Ping timeout: 248 seconds]
ScottHDev has quit [Quit: Ping timeout (120 seconds)]
ScottHDev has joined #nixos-dev
<evanjs>
eyJhb: didn't get too far with it. And the project I was looking at it that required it is now unmaintained/archived so 😀
aszlig has quit [Ping timeout: 246 seconds]
aszlig has joined #nixos-dev
<eyJhb>
Damn it :p I somewhat really want to get it to work, but at the same time I don't think it is actually required....
<eyJhb>
But how did you get it to build evanjs ?
<supersandro2000>
#118795 went to master by accident. reverting and doing a new PR to staging.
Raito_Bezarius has quit [Ping timeout: 246 seconds]
Raito_Bezarius has joined #nixos-dev
<abathur>
nix#4690 (3 word diff!) could use a look from someone with a handle on sudo security policy. It basically rewrites `sudo VAR=VAL command` -> `sudo env VAR=VAL command` to support `doas` as a sudo replacement. Since this pattern seems to let you leapfrog policy restrictions on env vars, I wonder if countermeasures are common?
<lukegb>
I think if you want to reasonably support policy restrictions on env vars you also have to lock down the commands you can execute
v0|d has joined #nixos-dev
Irenes has quit [Ping timeout: 248 seconds]
<abathur>
that's roughly my concern (new install failures driven by blanket blocking of env to prevent circumventing restrictions), though if the same setups almost always restrict HOME, there may not be much real compared to who would've already been seeing failures
<samueldr>
but I'm now curious to see if there's at least an order of magnitude approximation that can be derived from something
<samueldr>
though really I don't think we have any way of even doing that... but I'm not a statistician, and I don't know what we actually can know
justanotheruser has joined #nixos-dev
AtnNn has quit [Quit: Reconnecting]
AtnNn has joined #nixos-dev
cole-h has joined #nixos-dev
AtnNn has quit [Client Quit]
AtnNn has joined #nixos-dev
rajivr has quit [Quit: Connection closed for inactivity]
<siraben>
well what would we count exactly
<siraben>
because we can of course figure out how many times a package is depended upon
<siraben>
but should we count nix-env/nix-shell invocations, etc.? Not sure how we would tell them apart from regular derivations being built
<samueldr>
probably cannot do that because they won't go to the cache!
<samueldr>
(or just once)
<siraben>
right
<samueldr>
that's why I was reporting, knowing there's probably nothing that really can be done
<siraben>
I don't even know what we would gain from a measurement
<samueldr>
depends on what can be measured I guess :)
<siraben>
Since builds are reproducible, once or 100K times should be the same
<samueldr>
it'd count total number of substitutions from cache, if we're talking about cache hits
<siraben>
I'd be interested in seeing a report on most depended upon and broken (in hydra) packages
<samueldr>
but is it the only thing we have? I don't know
<siraben>
ah right
<samueldr>
and statistics is hard
<samueldr>
sometimes details are important, sometimes details are unimportant
<abathur>
hmm
pmy has quit [Ping timeout: 252 seconds]
<abathur>
if the goal is ballparking *users*, I guess disambiguating from infra is going to be a hurdle
<abathur>
the best proxies I can think of are nix/nixOS installer downloads over time, and nixpkgs tarball downloads? but AFAIK GH doesn't provide a way to check the latter
<samueldr>
users might not even be a metric that is wanted
<samueldr>
installation base? anything else?
<samueldr>
NixOS vs. Nix on non-NixOS could be interesting too
<samueldr>
and as I said I'm not a stats-person... what else is there that could be worthwhile?
<gchristensen>
google analytics in `nix`
<abathur>
they'd obviously launder a lot of infra, and miss a lot of re-use, but over time at least you'd know something relative
<gchristensen>
(puke)
<samueldr>
it's annoying when goals go directly against other goals :)
<aaronjanse>
I'm pretty sure Github shows tarball stats for releases and tags via their API
<abathur>
for that matter, though, total bandwidth over time probably says something fairly similar
<aaronjanse>
cache.nixos.org probably could also be used to estimate usage of various Nix things
<samueldr>
ballpart estimation, maybe more of change than absolute numbers
<samueldr>
ballpark*
<samueldr>
since overriding is so trivial with Nix it ends up being not-uncommon
<samueldr>
and then the fact it's a cache also obfuscates things a bit
<samueldr>
e.g. if you want install base, my machines all are built on a single machine, so generally I hit the cache once per update
<samueldr>
even for 3 machines
tazjin has quit [Quit: Byeee]
tazjin has joined #nixos-dev
tazjin has quit [Client Quit]
bpye has quit [Ping timeout: 268 seconds]
tazjin has joined #nixos-dev
<abathur>
aaronjanse: gh does have stats for the releases I gather, I'm just not sure if much of Nix's normal traffic flows through those; I guess the nix installer tarball downloads stats from releases.nixos.org, if they exist and are per path, might be useful
<abathur>
even if they are all-time, you'd be able to get some sense of change over time from the individual installers
<samueldr>
in fact I assume absolute numbers are mostly meaningless, but that over time might give insights
<abathur>
I guess from file dates that infra only dates back about a year, though, so I guess its perspective is limited unless earlier stats were also archived somewhere
<lukegb>
"huh, my home manager config doesn't work on macOS anymore" "oh, it's because I added ntfy to it and that has a darwin backend that needs pyobjc" "oh pyobjc is broken" "oh a bunch of macOS SDK frameworks are missing from nixpkgs" "oh pyobjc isn't really one package it's one per framework"
bpye0 has joined #nixos-dev
bpye has quit [Ping timeout: 268 seconds]
bpye0 is now known as bpye
justanotheruser has quit [Ping timeout: 248 seconds]
srk has quit [Ping timeout: 248 seconds]
justanotheruser has joined #nixos-dev
aminechikhaoui7 has joined #nixos-dev
fzakaria has quit [Quit: No Ping reply in 180 seconds.]
bgamari_ has joined #nixos-dev
blueberrypie6 has joined #nixos-dev
zhaofeng1 has joined #nixos-dev
fzakaria has joined #nixos-dev
zhaofeng has quit [Ping timeout: 268 seconds]
bgamari has quit [Ping timeout: 268 seconds]
aminechikhaoui has quit [Ping timeout: 268 seconds]
blueberrypie has quit [Ping timeout: 268 seconds]
blueberrypie6 is now known as blueberrypie
aminechikhaoui7 is now known as aminechikhaoui
bpye has quit [Quit: Ping timeout (120 seconds)]
bpye has joined #nixos-dev
bpye has quit [Ping timeout: 240 seconds]
bpye has joined #nixos-dev
bpye has quit [Quit: Ping timeout (120 seconds)]
bpye has joined #nixos-dev
<lukegb>
womp womp, cachix/cachix-action failure on nixpkgs because cloudflare cc domenkozar[m]
<lukegb>
oh wait, it was "520: Web server is returning an unknown error"