| 14 Feb 2025 |
 John Ericson | CI passed, the workaround for nix-eval-jobs worked! | 02:17:19 |
 Mic92 | Merged | 02:37:06 |
| John Ericson | https://github.com/nix-community/nix-eval-jobs/pull/353 Mic92 if you would be interested in having some sort of nix-next branch like I do for hydra, that would be nice | 03:37:39 |
| John Ericson | nbd if not (and going back and forth between upstream nix and nixpkgs nix is annoying) | 03:37:56 |
| John Ericson | otherwise a nix-next branch in hydra would have to point to a forked repo, which makes for a gross history | 03:38:30 |
| John Ericson | (I like to do things on Hydra and Nix master together, and then merge the branch in Hydra once Nix is released) | 03:38:49 |
|  sinan changed their profile picture. | 12:15:49 |
| 15 Feb 2025 |
|  BenjB83 joined the room. | 10:19:21 |
| BenjB83 changed their display name from Benjamín Buske to BenjB83. | 10:43:12 |
| 16 Feb 2025 |
|  Chris Hodapp joined the room. | 07:06:40 |
 Vladimír Čunát | The queue runner seems quite unstable now:
[root@mimas:~]# journalctl -u hydra-queue-runner.service -r | grep -F 'Main process exited'
Feb 16 09:16:02 mimas systemd[1]: hydra-queue-runner.service: Main process exited, code=dumped, status=11/SEGV
Feb 16 08:32:08 mimas systemd[1]: hydra-queue-runner.service: Main process exited, code=dumped, status=11/SEGV
Feb 16 06:26:30 mimas systemd[1]: hydra-queue-runner.service: Main process exited, code=dumped, status=11/SEGV
Feb 16 05:15:12 mimas systemd[1]: hydra-queue-runner.service: Main process exited, code=dumped, status=11/SEGV
Feb 16 04:44:47 mimas systemd[1]: hydra-queue-runner.service: Main process exited, code=dumped, status=11/SEGV
Feb 16 01:58:56 mimas systemd[1]: hydra-queue-runner.service: Main process exited, code=dumped, status=11/SEGV
Feb 15 17:48:36 mimas systemd[1]: hydra-queue-runner.service: Main process exited, code=dumped, status=11/SEGV
| 09:57:15 |
| Vladimír Čunát | The two traces with dump end with exactly the same gdb line and further down there are minimal differences.
(gdb) bt
#0 0x0000000000425fe6 in State::builder(std::shared_ptr<State::MachineReservation>) [clone .cold] ()
#1 0x0000000000493ac5 in std::thread::_State_impl<std::thread::_Invoker<std::tuple<void (State::*)(std::shared_ptr<State::MachineReservation>), State*, std::shared_ptr<State::MachineReservation> > > >::_M_run() ()
#2 0x00007f2b802be6d3 in execute_native_thread_routine () from /nix/store/4gk773fqcsv4fh2rfkhs9bgfih86fdq8-gcc-13.3.0-lib/lib/libstdc++.so.6
#3 0x00007f2b7ff5fd02 in start_thread (arg=<optimized out>) at pthread_create.c:447
#4 0x00007f2b7ffdf3ac in __GI___clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:78
| 10:05:49 |
| Vladimír Čunát | If I read it right, it's trying to dereference a NULL pointer with a 24B offset. | 10:17:23 |
 hexa | I think we've seen this particular trace for most of the last year 🫣 | 11:55:05 |
| hexa | wondering if its related to this change https://github.com/NixOS/hydra/commit/36e25d8fd2fe48132cffdf07ca0a564907758b21 | 11:57:05 |
| Vladimír Čunát | With this rate of crashes we'll be unable to finish builds like chromium. | 12:08:39 |
| Vladimír Čunát | (it never stays up for a few hours straight) | 12:09:03 |
| hexa | ok, will revert to test in a bit | 12:25:05 |
| hexa | or rather apply this https://git.lix.systems/the-distro/hydra/commit/fb9e29d4d0f2f591cd1d706fd3b7334af7d34b84 | 12:26:57 |
| hexa | applied | 12:29:35 |
| hexa | * deployed | 12:29:51 |
 K900 | Running a new eval on next | 12:32:07 |
| Vladimír Čunát | No crashes since then so far. A huge improvement 🎉 | 20:20:06 |
| 17 Feb 2025 |
| John Ericson | hexa (signing key rotation when): I have queued up a number of changes on master, and so I think it would be good to think about how to deploy | 00:24:08 |
| John Ericson | we could jump to the front and bisect, but I think probably walking linearly through the --first-parent history is more prudent | 00:24:31 |
| hexa | I think bumping to HEAD on my private hydra would be a start | 00:25:27 |
| hexa | error:
… while calling the 'attrNames' builtin
at «string»:1:207:
1| with builtins; let assets = getFlake "path:/tmp/nix-shell.sIGhyV/colmena-assets-EvihtE?lastModified=1739752075&narHash=sha256-yHR3T7GNYyN/b/GMe1oEW6j4S1EJqqfu9Hd/Me4imb4%3D"; hive = assets.processFlake; in attrNames hive.nodes
| ^
… while calling the 'getFlake' builtin
at «string»:1:29:
1| with builtins; let assets = getFlake "path:/tmp/nix-shell.sIGhyV/colmena-assets-EvihtE?lastModified=1739752075&narHash=sha256-yHR3T7GNYyN/b/GMe1oEW6j4S1EJqqfu9Hd/Me4imb4%3D"; hive = assets.processFlake; in attrNames hive.nodes
| ^
(stack trace truncated; use '--show-trace' to show the full, detailed trace)
error: cannot update unlocked flake input 'hive' in pure mode
[ERROR] -----
[ERROR] Operation failed with error: Child process exited with error code: 1
| 00:28:04 |
| hexa | cool, have to debug why nix doesn't like my colmena flake anymore first 🙂 | 00:28:17 |
| hexa | https://github.com/zhaofengli/colmena/issues/259 | 00:28:57 |
| hexa | ok | 00:46:54 |
