Nix Hackers | 893 Members | |
| For people hacking on the Nix package manager itself | 188 Servers |
| Sender | Message | Time |
|---|---|---|
| 18 Mar 2025 | ||
| it is unclear to me whether it is possible for two threads to be messing with a goal at once | 18:49:24 | |
| but if the goal is removed from worker, and from waitees/waiters, then I don't think another thread can find the goal very easily | 18:49:50 | |
| If Robert is busy today, well we at least have the Nix meeting tomorrow | 18:50:11 | |
| You can’t move around the worker interactions freely because you might mutate the data structures that are being iterated over at the same time | 18:50:38 | |
| Maybe not an issue there though | 18:50:48 | |
| Goals are single threaded | 18:51:12 | |
| Maybe in 2035 Mobile Element will support threads so I can respond it a thread | 18:52:03 | |
| * | 18:52:12 | |
| (Element X does) | 18:54:10 | |
| Normal Element does too | 18:54:27 | |
| Though we generally try to not use threads in public rooms because there's still clients that don't display them properly or at all | 18:54:45 | |
| I can’t make one on iOS I think? | 19:03:25 | |
| I never saw any mutex or something that made it obviously single threaded, but yes I believe that is the intent | 19:08:15 | |
| Sealed state was a mistake | 19:14:37 | |
| * | 21:56:23 | |
| * | 21:56:29 | |
| 19 Mar 2025 | ||
| 21:12:20 | ||
| 21:41:30 | |
| 20 Mar 2025 | ||
| Martin Schwaighofer: https://github.com/NixOS/nix/issues/12361 | 13:40:36 | |
| 17:52:14 | ||
| 17:52:43 | ||
| 17:54:05 | ||
| FYI in re https://git.lix.systems/lix-project/lix/issues/750, it seems like https://github.com/NixOS/nix/blob/3032512425a09fc58f2d658442043894e0aab256/src/libfetchers/tarball.cc#L74-L83 | 19:00:58 | |
| (and it does not seem like it can be trivially changed to recursive hash mode there, since that breaks a bunch of observable behaviour. i have no idea how this is supposed to behave, but it seems like a really quite large bug) | 19:01:44 | |
| * (and it does not seem like it can be trivially changed to recursive hash mode there, since that breaks a bunch of observable behaviour. i have no idea how this is supposed to behave, but it seems like a really quite tricky bug) | 19:01:50 | |
okay update: i can confirm that nixos/nix master emits that 9hagx path into the store and then additionally gets the q9prs one | 21:56:28 | |
* okay update: i can confirm that nixos/nix master emits that 9hagx path into the store and then additionally makes the q9prs one | 21:56:33 | |
| that seems at least inefficient if the files are large | 21:56:41 | |
| repro:
| 21:57:01 | |
| found some bonus context: https://github.com/NixOS/nix/pull/6548#discussion_r877921756 | 23:09:50 | |