| 4 Dec 2025 |
 helle (just a stray cat girl) | because ehgads | 17:44:08 |
 raitobezarius | open a CL with the changes | 18:17:40 |
|  ellie changed their display name from Ellie (The Fake One) to ellie. | 19:57:33 |
 9999years | hi. we have a nasty bug on at least some machines with at least lix 2.94.0 in some situations. definitely present on macos, not sure about linux | 20:04:16 |
| 9999years | the reproducer on my machine is this: cd $(mktemp -d) && nix flake init && nix flake lock --debug | 20:04:30 |
| 9999years | it gets stuck around here:
using cache entry '{"ingestionMethod":0,"name":"source","type":"file","url":"https://github.com/nixos/nixpkgs/archive/418468ac9527e799809c900eda37cbff999199b6.tar.gz"}' -> '{"etag":"W/\"5cd48868382e96938f2a04bff191018f8b59407d9ae34339c558b716c665372e\"","url":"https://codeload.github.com/NixOS/nixpkgs/tar.gz/418468ac9527e799809c900eda37cbff999199b6"}', '/nix/store/6zx3brkqjsnr53b2p3fdn2k44q6pyj25-source'
performing daemon worker op: 7
| 20:04:44 |
| 9999years | in some cases it may finish after a really long time (15 minutes) | 20:04:58 |
| 9999years | please run that command locally and let me know if it finishes. thanks! | 20:05:10 |
| raitobezarius | it finishes in 7s | 20:24:27 |
| raitobezarius | arguably 2.95 | 20:24:38 |
| raitobezarius | i can try to reproduce with 2.94.0 but idk if the daemon is an important piece or not | 20:24:49 |
| raitobezarius | that looks like a unexpected change in target URI | 20:25:21 |
| 9999years | wdym unexpected change in target uri? | 20:28:52 |
| 9999years | so it does finish eventually; 2.5 minutes for worker op 1 IsValidPath and 1.5 minutes for worker op 7 AddToStore:
Dec 04 12:30:06 performing daemon worker op: 1
Dec 04 12:30:06 using cache entry '{"ingestionMethod":0,"name":"source","type":"file","url":"https://github.com/nixos/nixpkgs/archive/418468ac9527e799809c900eda37cbff999199b6.tar.gz"}' -> '{"etag":"W/\"5cd48868382e96938f2a04bff191018f8b59407d9ae34339c558b716c665372e\"","url":"https://codeload.github.com/NixOS/nixpkgs/tar.gz/418468ac9527e799809c900eda37cbff999199b6"}', '/nix/store/6zx3brkqjsnr53b2p3fdn2k44q6pyj25-source'
Dec 04 12:32:33 finished performing daemon worker op: 1
Dec 04 12:32:33 performing daemon worker op: 7
Dec 04 12:33:58 finished performing daemon worker op: 7
| 20:35:21 |
| 9999years | the 1.5 minute times are pretty predictable -- we got any 90 second timeouts hanging around in there? 😬 | 20:47:25 |
| raitobezarius | github changing URI after the first GET | 21:41:07 |
| raitobezarius | it'd be helpful to have a strace on the Darwin daemon or more logging | 21:41:26 |
| raitobezarius | is this with maximum logging? | 21:41:31 |
| raitobezarius | we have various timeouts mechanisms at the curl level for DNS resolution and so on | 21:42:28 |
| raitobezarius | maybe your DNS resolution is broken? | 21:42:32 |
| 9999years | --debug, similar results with --debug -vvv | 21:55:32 |
| 9999years | seems unlikely but not impossible | 21:55:46 |
| raitobezarius | 9999years cannot reproduce on macOS arm64 | 22:28:35 |
| raitobezarius | uhm i'm stupid | 22:28:44 |
| raitobezarius | this is running 2.93.3 | 22:28:46 |
| raitobezarius | let's upgrade the macOS fleet | 22:29:02 |
| raitobezarius | https://cl.afnix.fr/c/infra/+/174 | 22:30:36 |
| raitobezarius | we will see in the next minutes when the MDM redeploys the machine whether this is broken or not | 22:30:56 |
| 9999years | we've got at least intermittent reports of it on 2.93.3 as well | 23:23:07 |
| 9999years | i don't think the bug is super deterministic, whatever it is | 23:23:20 |
