gchristensen changed the topic of #nixos-borg to: https://www.patreon.com/ofborg https://monitoring.nix.ci/dashboard/db/ofborg?refresh=10s&orgId=1&from=now-1h&to=now "I get to skip reviewing the PHP code and just wait until it is rewritten in something sane, like POSIX shell. || https://logs.nix.samueldr.com/nixos-borg
hmpffff has quit [Quit: nchrrrr…]
hmpffff has joined #nixos-borg
hmpffff has quit [Quit: nchrrrr…]
orivej has joined #nixos-borg
cole-h has quit [Quit: Goodbye]
tilpner has quit [Quit: tilpner]
tilpner has joined #nixos-borg
<LnL> urgh, enumerating nixpkgs properly requires recursion detection
<srk> :))
<srk> how come?
orivej has quit [Ping timeout: 264 seconds]
<srk> LnL: what I'm trying to do is to enumerate the same way as hydra-eval-jobs from release-combined.nix but with hnix
<srk> I also have a PoC of AST traversal for following callPackage which works for like ~9k packages
<srk> and a combinator setBreakage FilePath Bool which can adjust nix files (add/replace/remove broken) similar to what update-nix-fetchgit does with src/name
<srk> I will generalize this to arbitrary nix file modification operations to be able to e.g. add attr to set according to path or filter attributes from set
<LnL> well stuff hydra doesn't enumerate should never be removed?
<LnL> anyway, this thing is fine to start with
<srk> it should of course but that's just a matter of altering allowBroken allowUnfree and building yourself
<srk> and recurseIntoAttrs default I think
<LnL> my point was that recurseIntoAttrs is whitelist and you kind of want a blacklist instead
<srk> which part is problematic recursion wise?
<LnL> things like pkgs, __splicedPackages, etc.
<srk> ah, indeed there's pkgs.pkgs..
<srk> cool, that's very useful
<srk> LnL++
<{^_^}> LnL's karma got increased to 34
<LnL> that gives us {"broken":false,"insecure":false,"pkg":"zzuf","unfree":false,"unsupported":false}
<srk> I was confused about .available and why checkMeta is hidden like this
<LnL> which means this + the channel revision/eval it ran against is should do the trick to analyze most of the jobs in hydra
<srk> yup
<srk> I've played with tryEval and looking for meta but it wasn't that nice
orivej has joined #nixos-borg
hmpffff has joined #nixos-borg
hmpffff has quit [Quit: nchrrrr…]
hmpffff has joined #nixos-borg
WilliButz has quit [Quit: bye]
WilliButz has joined #nixos-borg
<LnL> gchristensen: ping
<gchristensen> pong, but in about 10minutes I'll be AFK for a fetw hours
<LnL> nothing urgent, but I was wondering if I could get a partial copy of the hydra data
<LnL> would be nice to have some real data
<gchristensen> you bet
<gchristensen> but not for a few hours :')
<LnL> sure, but I'll prepare some commands then
<gchristensen> cool
<gchristensen> I'm thinking I'll get you everything but the users table
<LnL> don't really need everything and it's a bit big otherwise
<gchristensen> do you want subsets of tables, or individual tables?
<LnL> something like that gives the data for nixpkgs/trunk over the last year, which should be more than enough for now
hmpffff has quit [Quit: Bye…]
hmpffff has joined #nixos-borg
hmpffff_ has joined #nixos-borg
hmpffff has quit [Ping timeout: 256 seconds]
hmpffff_ has quit [Quit: nchrrrr…]
cole-h has joined #nixos-borg
<cole-h> Is it wrong that I *want* GH to break their API again to see if alerts are working...? :D
<gchristensen> :D
<qyliss> cole-h: careful what you wish for ;)
<MichaelRaskin> Maybe the secret wish is to have a full Nix-based CI for ofborg, and GitHub breaking API is the way of forcing the move to something reliably CI-able?
<gchristensen> =)
cole-h has quit [Quit: Goodbye]
cole-h has joined #nixos-borg
<MichaelRaskin> I wonder if it is possible to break ofborg by a bad command
<MichaelRaskin> Purely for alert testing, of course
<gchristensen> give it a go :)
<MichaelRaskin> My natural guessses have a drawback of maybe succeeding and queueing a ton of builds
<MichaelRaskin> Or taking a forever to fail to evaluate
<cole-h> Am I the only one that thinks the 501-5000 rebuild tags being white (or off-white) is weird? 11-100 is yellow, 101-500 is orange, 501+/5001+ are red...
<cole-h> I guess only tangentially related to ofborg, but still
<gchristensen> nobody has bothered to give them a color I guess
orivej has quit [Ping timeout: 256 seconds]