 | Lix Development | 429 Members |
| (Technical) development of Lix, the package manager, a Nix implementation. Please be mindful of ongoing technical conversations in this channel. | 141 Servers |
| Lix Development | 429 Members |
| (Technical) development of Lix, the package manager, a Nix implementation. Please be mindful of ongoing technical conversations in this channel. | 141 Servers |
| Sender | Message | Time |
|---|---|---|
| 21 Apr 2026 | ||
 zoë (@blokyk) | * with regards to #1175, does saving each typed line ourselves (e.g. in an std::list or std::vector?) and then appending it to the file sound good? if so i can try to make a CL for it, but i'd like to get a vibe-check for it first | 12:38:30 |
| zoë (@blokyk) | * with regards to #1175, does saving each typed line ourselves (e.g. in an std::list or std::vector?) and then appending it to the file sound good? if so i can try to make a CL for it, but i'd like to get a vibe-check about it first | 12:38:35 |
 piegames | Hm, would simply saving the history after each command also work around the issue? | 12:46:25 |
| piegames | (Not sure if that's a good idea either) | 12:46:41 |
| zoë (@blokyk) | no unfortunately, unless we also reload the history before each prompt, because otherwise we'll still be writing the "outdated" history to the file | 12:48:12 |
| zoë (@blokyk) | for most people it shouldn't be an enormous burden, but idk it feels dirty and pretty fragile | 12:50:44 |
| piegames | In reply to @blokyk:matrix.orgThat's what I meant yes | 12:55:12 |
| zoë (@blokyk) | * for most people it shouldn't be an enormous burden, but idk it feels dirty and pretty fragile (we'd be at the mercy of disk I/O timings) | 12:56:10 |
| zoë (@blokyk) | * for most people it shouldn't be an enormous burden, but idk it feels dirty and pretty fragile (+we'd be at the mercy of disk I/O timings) | 12:56:21 |
| zoë (@blokyk) | personally i don't feel like that's a better solution :/ in part because of the aforementioned fragility+timing, but also because it'd then have the side effect of sharing the history between every repl session, which is an antifeature imo (i often have two repl sessions side-by-side to trying out different things, or i open a quick repl to check something for a bigger expr or something; if both sessions are shared i have to wade through the other repl's history while trying to navigate the current session's history because they're now all confusingly interleaved... there's a reason i generally disable zsh's SHARE_HISTORY option @_@) | 13:04:06 |
| piegames | I see | 13:11:14 |
 goldstein | offtopic, but thank you for teaching me I can just disable SHARE_HISTORY, this was annoying me immensely | 14:46:20 |
| zoë (@blokyk) | there's actually multiple mutually-exclusive settings that control zsh, each one making you go "wait, the other one didn't do that?!" :D
| 14:51:54 |
| 23 Apr 2026 | ||
 antifuchs | also can I just say log-format = multiline-with-logs is so nice! first time I'm customizing that log format and it is far easier to see what's going on during nix build | 17:29:53 |
| antifuchs | (the explanation in the manual is a little less clear re. the differences to bar-with-logs though. is the diff that multiline-with-logs shows the currently active tasks below the bar?) | 17:31:09 |
 kloenk | yeah essentially. sorry I'm bad with describing some things :) | 17:33:59 |
| antifuchs | On a re-read that’s what it says! Just hard to form a mental picture without seeing it in action tbh | 19:56:35 |
| 25 Apr 2026 | ||
 hsjobeki joined the room. | 11:19:23 | |
| hsjobeki | I'm planning to migrate the nixpkgs module system eval tests from bash to nix-unit. There is also lix-unit. Question in the room: any plans to add this to nixpkgs? I would then wire up the test suite against that. Or do people have there own module system tests in lix? | 11:21:53 |
 emily | does nix-unit support all the supported versions of the Nix C++ API? otherwise we lose the ability to test on those versions, no? | 11:29:27 |
| emily | and https://github.com/adisbladis/lix-unit is archived for years unless it was moved | 11:30:20 |
 Sergei Zimmerman (xokdvium) | In reply to @hsjobeki:matrix.orgI’m not sure it’s a great idea. The tests are used for both cppnix/lix CI I think, and it would be quite intractable. Maybe if it could be polyfilled somehow then it would be nice yes | 11:36:44 |
| hsjobeki | Hm. Trying to get rid of this bash script for years now... Maybe there is a different way. | 12:14:28 |
| 27 Apr 2026 | ||
| zoë (@blokyk) | i'm not sure why, but the repl-characterization-tests are not failing as expected in some cases.for example, if i change one of the data files (e.g. replacing "hello" with "hi" in tests/functiona/repl_characterization/data/regression_9917.nix), the test doesn't fail when it clearly should.(i'm noticing this because i'm trying to understand why tests didn't fail in cl/5491 when they clearly should have) | 11:04:40 |
| zoë (@blokyk) | * i'm not sure why, but the repl-characterization-tests are not failing as expected in some cases.for example, if i change one of the data files (e.g. replacing "hello" with "hi" in tests/functiona/repl_characterization/data/regression_9917.nix), the test doesn't fail when it clearly should.edit: even after a clean rebuild, it just doesn't seem to pick up on the change, which is... so strange @_@ (i'm noticing this because i'm trying to understand why tests didn't fail in cl/5491 when they clearly should have) | 11:22:37 |
| zoë (@blokyk) | ok even changing the actual test (not just the data) still doesn't make the test fail, so i'm not sure what is even happening. | 11:36:34 |
| zoë (@blokyk) | * ok even changing the actual test (not just the data) still doesn't make the test fail, so i'm not sure what is even happening. | 11:36:53 |
| zoë (@blokyk) | i'm gonna try to bisect this, but i don't have a very powerful machine (and i'm kinda doing this as a side-activity while waiting for other stuff) | 11:38:15 |
| zoë (@blokyk) | * i'm gonna try to bisect this, but i don't have a very powerful machine (and i'm kinda doing this as a side-activity in-between two llvm builds) | 11:38:28 |
 flokli | Poke, any opinion on https://git.lix.systems/lix-project/lix/issues/1185 ? It would be nice if this were consistent across Lix and Cppnix, we ran into this in the snix test suite. | 14:27:52 |