| 31 Aug 2025 |
|  teutat3s joined the room. | 16:44:40 |
| teutat3s | Is there any workaround to fix a cached failure on hydra that cannot be reproduced locally? Or is the solution to just wait until some cache expiry?
Asking for https://hydra.nixos.org/build/306350994 | 16:47:29 |
 K900 | Restarted | 17:05:05 |
| 1 Sep 2025 |
 Mic92 | John Ericson: Since we switched to nix-eval-jobs in hydra, I noticed that evaluation is no longer "pure" because it uses --expr in combination with builtins.getFlake. So I added this pull request to have an alternative map function that can be applied to the expression that nix-eval-jobs receives: https://github.com/nix-community/nix-eval-jobs/pull/378 | 13:47:52 |
 John Ericson | @joerg:thalheim.io: I think it can still be pure? | 15:47:58 |
| John Ericson | Can get flake with lock file info be done with pure eval? | 15:48:29 |
| Mic92 | https://github.com/nix-community/nix-eval-jobs/blob/a579b1a416dc04d50c0dc2832e9da24b0d08dbac/src/nix-eval-jobs.cc#L454 | 15:49:14 |
| John Ericson | If you pass --pure-eval what happens? | 15:51:51 |
| John Ericson | I thought we had some logic to do an initial fetch so we had the info we needed for pure eval? | 15:52:11 |
| John Ericson | We can do the new flag in nix eval jobs, but it isn't really a nix eval jobs -specific problem, so I rather have a more general solution | 15:52:46 |
| Mic92 | John Ericson: that wouldn't work with getFlakes because those inputs are not populated into the nix path. | 16:15:04 |
| Mic92 | * John Ericson: --pure-eval- wouldn't work with getFlakes because those inputs are not populated into the nix path. | 16:25:01 |
| Mic92 | * John Ericson: --pure-eval wouldn't work with getFlakes because those inputs are not populated into the nix path. | 16:25:06 |
|  Aangularity joined the room. | 17:31:19 |
| 2 Sep 2025 |
|  Wolfgang Walther joined the room. | 14:09:19 |
| 3 Sep 2025 |
|  ghpzin joined the room. | 07:30:01 |
| 4 Sep 2025 |
|  connor (he/him) (UTC-7) changed their display name from connor (he/him) (UTC+1) to connor (he/him) (UTC+2). | 17:02:53 |
| 8 Sep 2025 |
|  inayet set a profile picture. | 02:15:40 |
| 9 Sep 2025 |
|  aciceri changed their display name from zrsk to aciceri. | 15:02:14 |
| 10 Sep 2025 |
| connor (he/him) (UTC-7) changed their display name from connor (he/him) (UTC+2) to connor (he/him) (UTC-7). | 22:20:24 |
| 11 Sep 2025 |
|  loudgolem joined the room. | 17:50:59 |
|  bumperboat joined the room. | 23:24:01 |
| bumperboat | there was a kernel regression causing tailscale test to fail, which was supposedly resolved in 6.12.46. how to see if the hydra build servers are running a fixed kernel to pass the test and what would trigger an attempt to rebuild/test tailscale? | 23:26:44 |
| 12 Sep 2025 |
 dramforever | relevant folks are in -> #infra:nixos.org | 00:27:52 |
|  @creepinson:matrix.org left the room. | 01:43:10 |
| 14 Sep 2025 |
|  Emma [it/its] joined the room. | 08:39:32 |
| 16 Sep 2025 |
 shawn8901 | Hi,
Could it be, that the Github Status Plugin is currently not working as expected? Since some time i am missing the reported build status for my flake on github.
The error message according hydra-notify is the following:
Sep 16 18:22:56 tank hydra-notify[2039]: Flake is github:shawn8901/nixos-configuration/28dbe4cc7f9b036bb7a3fc37a22b08730f48df0d?narHash=sha256-BeuUHKxtlcnweo9c2rZLRABgIv2ARBUXfySQ6MJDxQA%3D
Sep 16 18:22:56 tank hydra-notify[2039]: Can't parse flake, skipping GitHub status update
https://github.com/NixOS/hydra/blob/274027eb504c7fe090e00c16fd94f4b832981095/src/lib/Hydra/Plugin/GithubStatus.pm#L100
Which is right according the regex, tho i am not seeing what should be wrong on my end, it looks to me that the narHash that is in the url is unexpected.
My jobset definition is here: https://github.com/Shawn8901/nixos-configuration/blob/b1f6e066c77c4eb078bcf873bf6e7eeb3a91db16/.hydra/jobsets.nix#L28
And i dont see where i could cause the narHash to be included and i asume its more a internal issue here.
| 19:18:47 |
| shawn8901 | * Hi,
Could it be, that the Github Status Plugin is currently not working as expected for flakes? Since some time i am missing the reported build status for my flake on github.
The error message according hydra-notify is the following:
Sep 16 18:22:56 tank hydra-notify[2039]: Flake is github:shawn8901/nixos-configuration/28dbe4cc7f9b036bb7a3fc37a22b08730f48df0d?narHash=sha256-BeuUHKxtlcnweo9c2rZLRABgIv2ARBUXfySQ6MJDxQA%3D
Sep 16 18:22:56 tank hydra-notify[2039]: Can't parse flake, skipping GitHub status update
https://github.com/NixOS/hydra/blob/274027eb504c7fe090e00c16fd94f4b832981095/src/lib/Hydra/Plugin/GithubStatus.pm#L100
Which is right according the regex, tho i am not seeing what should be wrong on my end, it looks to me that the narHash that is in the url is unexpected.
My jobset definition is here: https://github.com/Shawn8901/nixos-configuration/blob/b1f6e066c77c4eb078bcf873bf6e7eeb3a91db16/.hydra/jobsets.nix#L28
And i dont see where i could cause the narHash to be included and i asume its more a internal issue here.
| 19:19:18 |
| shawn8901 | I saw that there is https://github.com/NixOS/hydra/issues/1486 but my hydra.conf looks fine according to what is mentioned in the issue
❯ cat /var/lib/hydra/hydra.conf
evaluator_max_memory_size = 4096
evaluator_workers = 4
max_concurrent_evals = 1
restrict-eval = false
max_output_size = 5368709120
max_db_connections = 150
compress_build_logs = 1
<runcommand>
job = *:*:merge-pr
command = /nix/store/vfw63pq1i2r937j7aszff38msfi4d0y8-merge_pr/bin/merge_pr
</runcommand>
<hydra_notify>
<prometheus>
listen_address = 127.0.0.1
port = 9199
</prometheus>
</hydra_notify>
<githubstatus>
jobs = .*
useShortContext = true
</githubstatus>
Include /run/secrets/rendered/hydra-write-token.conf
Include /run/secrets/rendered/hydra-hook-token.conf
using_frontend_proxy = 1
base_uri = https://hydra.pointjig.de
notification_sender = hydra@pointjig.de
max_servers = 25
gc_roots_dir = /nix/var/nix/gcroots/hydra
use-substitutes = 1
| 19:23:16 |
| shawn8901 | * I saw that there is https://github.com/NixOS/hydra/issues/1486 but my hydra.conf looks fine according to what is mentioned in the issue
❯ cat /var/lib/hydra/hydra.conf
evaluator_max_memory_size = 4096
evaluator_workers = 4
max_concurrent_evals = 1
restrict-eval = false
max_output_size = 5368709120
max_db_connections = 150
compress_build_logs = 1
<runcommand>
job = *:*:merge-pr
command = /nix/store/vfw63pq1i2r937j7aszff38msfi4d0y8-merge_pr/bin/merge_pr
</runcommand>
<hydra_notify>
<prometheus>
listen_address = 127.0.0.1
port = 9199
</prometheus>
</hydra_notify>
<githubstatus>
jobs = .*
useShortContext = true
</githubstatus>
Include /run/secrets/rendered/hydra-write-token.conf
Include /run/secrets/rendered/hydra-hook-token.conf
using_frontend_proxy = 1
base_uri = https://hydra.pointjig.de
notification_sender = hydra@pointjig.de
max_servers = 25
gc_roots_dir = /nix/var/nix/gcroots/hydra
use-substitutes = 1
so the config part of the issue seems unrelated and its more the 2nd part about the url.
I assume issue author did later just use the git url variant, but i dont think that this is a proper solution as the github url is quite nice to use.
| 19:25:24 |
