| 10 Aug 2022 |
 Amanda (she/her) | shame, oh well | 18:32:24 |
 ajs124 | it does? | 18:32:32 |
| ajs124 | but it substitutes from multiple different caches for just fine, for one closure | 18:32:51 |
| ajs124 | * but it substitutes from multiple different caches just fine, for one closure | 18:33:02 |
 @grahamc:nixos.org | I mean, maybe it does actually do that in reality, but the invariant / expectation is still there | 18:33:27 |
| ajs124 | good to know 😬 | 18:33:43 |
| Amanda (she/her) | Hum, that may actually prove useful, since my local cache's growing pretty big. 1.1TiB now | 19:02:59 |
| Amanda (she/her) | So I can just nuke the directory without worrying about dangling references | 19:03:14 |
| Amanda (she/her) | ( I assume ) | 19:03:20 |
| Amanda (she/her) | Unless there's another way to prune it, last I heard there's no such tools for that publically | 19:03:50 |
| ajs124 | I think I saw one in a gist somewhere, for s3 caches | 19:07:45 |
| ajs124 | https://gist.github.com/DarkKirb/533e9e8b9b6dd1badf9b103b4285dc4e | 19:08:06 |
| Amanda (she/her) | interesting | 19:09:48 |
| ajs124 | I also hacked one up, but it's very dumb and just works on a directory of narinfo and nar files by basically doing LRU.
Doesn't even try to parse any dependencies or anything rn. | 19:10:50 |
|  kenji joined the room. | 19:47:30 |
| 14 Aug 2022 |
|  @rimuru:gentoo.chat changed their profile picture. | 09:36:44 |
 @linus:schreibt.jetzt | I wrote https://github.com/lheckemann/cache-gc which operates on a flat-file binary cache and removes things based on mtime (and keeps dependencies of "live" things even if they're older) | 10:56:57 |
| @linus:schreibt.jetzt | ajs124: how do you get LRU info? | 10:57:16 |
| @linus:schreibt.jetzt | access logs? | 10:59:44 |
 das_j | In reply to @linus:schreibt.jetzt
ajs124: how do you get LRU info? Atime iirc | 11:00:28 |
| @linus:schreibt.jetzt | aah ok | 11:00:38 |
| @linus:schreibt.jetzt | I disable that everywhere 😅 | 11:00:52 |
| @linus:schreibt.jetzt | you could use mine with a one-byte change to preserve the dependency invariant :p | 11:02:12 |
| das_j | Yeah we explicitly enabled it on that Filesystem | 11:02:16 |
| das_j | Probably the only ones with atime in our setup | 11:02:27 |
| @linus:schreibt.jetzt | unrelated, I'm trying to set up a new project in my hydra and getting this from the evaluator:
hydra-eval-jobs returned exit code 1:
Collecting from unknown thread
child process (251757) killed by signal=6
error: unexpected EOF reading a line
| 11:03:12 |
| @linus:schreibt.jetzt | what's the most helpful way to get more useful info about what's going wrong? | 11:03:26 |
| @linus:schreibt.jetzt | (I'm 90% sure this is a nix bug and not my fault, but...) | 11:03:37 |
| das_j | You can maybe check coredumpctl? | 11:05:02 |
