| 21 Jan 2026 |
 raitobezarius | https://source.android.com/docs/core/runtime/zygote | 22:41:52 |
| raitobezarius | i figured out another piece of docs to explain the problem than the obscure piece of docs of zygote of Android | 23:51:28 |
| raitobezarius | https://unikraft.org/blog/2025-05-15-multiprocess#posix-and-the-challenges-of-multiprocess | 23:51:29 |
| raitobezarius | another plan we have to solve this is to be able to use vfork | 23:51:40 |
| raitobezarius | (the path is being slowly opened by the recent chains on reworking libexec helpers in https://gerrit.lix.systems/c/lix/+/4994) | 23:51:56 |
| raitobezarius | the important piece is | 23:52:42 |
| raitobezarius |
given that before COW fork() was too slow as the kernel had to copy the parent's context.
| 23:52:44 |
| 22 Jan 2026 |
 antifuchs | Ah, the classic fork-copies-everything issue, eh. Thanks for linking that! | 01:00:45 |
 Rutile (rootile) | https://lix.dev/blog/introducing-lix/
Ohno | 08:29:06 |
 hexa | https://youtu.be/t6-cCh9bTG4?si=eH4hZE-9YZaGuCO-&t=14 | 08:40:43 |
 m | In reply to @commentator2.0:elia.garden
https://lix.dev/blog/introducing-lix/
Ohno #molotov | 15:21:38 |
 Acid Bong | In reply to @commentator2.0:elia.garden
https://lix.dev/blog/introducing-lix/
Ohno i guess Lix isn't registered yet | 16:09:43 |
| Acid Bong | (the bane) | 16:09:52 |
| Acid Bong | *name | 16:09:58 |
|  Lotte (it/its)/Cinny (she/her) θΔ& changed their profile picture. | 16:47:31 |
| antifuchs | pretty sure there are witches working on the lix project that could arrange for a bane | 17:13:09 |
|  pentane changed their profile picture. | 22:21:03 |
| 23 Jan 2026 |
|  @io:meat.computer left the room. | 00:52:51 |
 niko ⚡️ | Does Lix expose any probes for profiling, especially when it comes to eval? | 11:51:38 |
 piegames | What do you mean with probes? If you want to profile your Nix code then it has performance counters you can print with some flags (IIRC), when it comes to profiling eval itself then perf and callgrind it is | 12:44:10 |
 thubrecht | NIX_SHOW_STATS=1 is the env variable that makes it print the counters | 12:54:39 |
|  nea joined the room. | 13:05:52 |
| raitobezarius | we have USDT probes for HTTP transfers | 15:24:01 |
| raitobezarius | i want to introduce more probes, like, significantly more | 15:24:09 |
| raitobezarius | so that it's possible to write an eBPF profiler based on them for eval | 15:24:18 |
| raitobezarius | but didn't have time | 15:24:26 |
| niko ⚡️ | In reply to @piegames:flausch.social
What do you mean with probes? If you want to profile your Nix code then it has performance counters you can print with some flags (IIRC), when it comes to profiling eval itself then perf and callgrind it is I mean not profiling the eval implementation itself but profiling nix code against the current eval implementation. But those answer my question, thank you, I’ll see if the counters are sufficient for what I want to look into. Worst case I’ll just wait with anticipation for raito to get back to introducing more probes | 15:54:23 |
| raitobezarius | there's a function tracer as well fyi | 16:38:47 |
|  vczf changed their profile picture. | 18:47:02 |
| 24 Jan 2026 |
 Milo | lix fmt doesn't behave the same as nix fmt when invoked with no arguments.
nix fmt passes the project root directory (the one parent of the flake.nix) as the default argument.
Whereas, lix fmt simply acts as a nix run .#formatter.
Is this intentional? If so, what is the reason? If not, could I volunteer a PR to match nix's behaviour?
| 00:13:35 |
