| 17 Jun 2026 |
 K900 | Well, /nix/sto | 13:52:44 |
| K900 | Technically we can't prove what comes next | 13:52:48 |
| K900 | But I am fairly certain | 13:52:57 |
 hexa | can someone remind me, how was the old queue-runner. did it complete builds and whether or not the upload failed did not affect the job state? | 14:10:01 |
| hexa | and maybe we're now getting much more visibility into upload problems, because upload problems abort builds. | 14:10:29 |
| K900 | We had aborted builds due to upload failures before | 14:11:28 |
| K900 | I'm pretty sure this isn't a real upload failure | 14:11:42 |
| K900 | But whatever that parse error is | 14:11:42 |
| hexa | yeah, I'd like a more convincing error message than
No discriminant in enum RawLogMessageType matches the value 8031170683229269551
| 14:13:32 |
| K900 | ...I have a theory actually | 14:14:30 |
| K900 | What are the odds it's trying to parse the log protocol | 14:14:45 |
| K900 | And instead it just gets nothing and then a store path | 14:14:50 |
| K900 | Because the store path already exists | 14:15:00 |
| hexa | https://github.com/NixOS/hydra/blob/f69a55cf3c73146ba4ac4d5e7d39377a226d85fb/subprojects/hydra-builder/src/state.rs#L629-L639 upload_nars called here | 14:19:26 |
| hexa | and I don't see any retry logic | 14:20:09 |
| hexa | not in upload_nars, not in upload_nars_presigned | 14:20:39 |
| hexa | we deployed another update to fix grpc connection aborts from nginx | 16:46:27 |
| K900 | https://hydra.nixos.org/build/331658388 | 18:03:29 |
| K900 | Another unsupported system situation | 18:03:33 |
| hexa | 
Download | 23:22:23 |
| hexa | uhhhhhhh | 23:22:27 |
| hexa | 160k rebuilds seems a bit excessive here | 23:22:58 |
| hexa | ❯ nix-diff /nix/store/m74651b793zgyyvlk9gx7v1cl1ywslib-hello-2.12.3.drv /nix/store/gh9rl6g6jw0kvkzi3mh40ig2phffrwjz-hello-2.12.3.drv
- /nix/store/m74651b793zgyyvlk9gx7v1cl1ywslib-hello-2.12.3.drv:{out}
+ /nix/store/gh9rl6g6jw0kvkzi3mh40ig2phffrwjz-hello-2.12.3.drv:{out}
• The input derivation named `hello-2.12.3.tar.gz` differs
- /nix/store/jk8228a88vjsa0zk004sgdhxp5g4pnnv-hello-2.12.3.tar.gz.drv:{out}
+ /nix/store/b1nwvxqz0prj6kqvwfs7i8d7sk40m4w7-hello-2.12.3.tar.gz.drv:{out}
• The input derivation named `mirrors-list` differs
- /nix/store/84w0zzmqzl2bn5s4rx7axsh9znws1rdf-mirrors-list.drv:{out}
+ /nix/store/qkrfwhcrn7af1w11gxh3m45xjsgxlrxz-mirrors-list.drv:{out}
• The environments do not match:
+ texhistoric=https://ftp.math.utah.edu/pub/tex/historic/ https://texlive.info/historic/ https://ftp.tu-chemnitz.de/pub/tug/historic/ https://pi.kwarc.info/historic/ https://mirrors.tuna.tsinghua.edu.cn/tex-historic-archive/ https://mirror.nju.edu.cn/tex-historic/ ftp://tug.org/texlive/historic/
• Skipping environment comparison
• Skipping environment comparison
| 23:26:37 |
| hexa | mirror list change looks sus | 23:26:41 |
| hexa | https://github.com/NixOS/nixpkgs/pull/518480 | 23:27:53 |
| hexa | yup, 77456b834919be3e6e8ed00bd26eebd32046dc8e causes a rebuild for hello | 23:31:20 |
| hexa | heh | 23:32:39 |
| hexa | 
Download | 23:32:47 |
| 18 Jun 2026 |
| hexa | ok, trying out iocaine instead of anubis on hydra | 00:30:19 |
| hexa | if you see random markov shit instead of hydra, feel free to reach out | 00:30:32 |
