 | Lix Development | 346 Members |
| (Technical) development of Lix, the package manager, a Nix implementation. Please be mindful of ongoing technical conversations in this channel. | 125 Servers |
| Lix Development | 346 Members |
| (Technical) development of Lix, the package manager, a Nix implementation. Please be mindful of ongoing technical conversations in this channel. | 125 Servers |
| Sender | Message | Time |
|---|---|---|
| 26 Apr 2025 | ||
 Winter | In reply to @jade_:matrix.orgI opened a radar about that, let me check on it | 06:02:32 |
| Winter | like years ago | 06:02:37 |
| Winter |
| 06:03:13 |
| Winter | lmao | 06:03:15 |
 jade_ | re frustrations with pytest, those still exist, but i don't think it is necessarily that big of a hassle to port the test suite to a different one when we have the time to write a better one. the fixtures in pytest are almost good, the only thing i want different from them is that they should be driven directly by type annotations the big bad thing about pytest is xdist related jankiness, but i also doubt it is going to wind up ultimately being That Bad. the ideal testsuite would have threaded or asyncio running instead of the xdist multiprocessing monster, but it's kinda like whatever. | 06:03:16 |
| jade_ | * re frustrations with pytest, those still exist, but i don't think it is necessarily that big of a hassle to port the test suite to a different one when we have the time to write a better one. the fixtures in pytest are almost good, the only thing i want different from them is that they should be driven directly by type annotations rather than names the big bad thing about pytest is xdist related jankiness, but i also doubt it is going to wind up ultimately being That Bad. the ideal testsuite would have threaded or asyncio running instead of the xdist multiprocessing monster, but it's kinda like whatever. | 06:03:36 |
| Winter | In reply to @jade_:matrix.org i’m also not 100% sure if this just applies to the server or not. i’d imagine it applies to both. but also wait, i’m out of the loop: we already use the system/PATH’s ssh binary for stuff, no? | 06:06:09 |
| jade_ | oh i mean in the case of using libssh in place of ssh | 06:06:25 |
| jade_ | that this would potentially raise some issues with that. and that those are -fine- but we just have to be aware of them and deliberate about communicating since it is likely to regress some unusual use cases | 06:07:05 |
| Winter | In reply to @jade_:matrix.orgOH you mean because of keychain. | 06:07:44 |
| jade_ | yup | 06:07:52 |
| jade_ | i was a bit unclear sorry | 06:08:01 |
| jade_ | me? nope i don't use it system wide because it breaks too much stuff haha | 06:08:58 |
| jade_ | * me? nope i don't use it system wide because after my run-in with fish in 2016 i decided non posix shells break too much random stuff haha | 06:09:24 |
| Winter | do you use it interactively? :thinking: | 06:09:28 |
| Winter | lol | 06:09:31 |
| jade_ | only when working on releng | 06:09:36 |
 K900 | Honestly probably a good call | 06:09:37 |
| Winter | oh so the xsh scripts in releng are xonsh?! is there not a shebang there on purpose? | 06:09:49 |
| jade_ | it would be a nice replacement for ipython, i should actually um. use it for that. | 06:09:51 |
| jade_ | yes that is on purpose! | 06:10:00 |
| jade_ | they are imported! | 06:10:03 |
| jade_ | it's a multi-language python project, really | 06:10:22 |
| jade_ | the xonsh bits are where xonsh is a better fit, and the python bits are where i want a type checker | 06:10:47 |
| Winter | i… see… | 06:13:09 |
| Winter | very interesting | 06:13:13 |
| Winter | i did not know that was possible! | 06:13:19 |
| jade_ | like, in re this, i do not know what are the current typical patterns in python since i write exactly enough of it to be pretty good at the base language and not know much about the library ecosystem or packaging or docstring standards. so i have basically no opinions (yet i wrote a lot of the python around in lix). the more important thing is that we have a standard, rather than what it is. so whatever gives the nicest tooling support and is the modern standard is Fine. actually 9999years do you know things about this? googling around has found me "docments", which just puts the param documentation as a comment above the param, the sphinx thing, PEP-727 which puts things into the type annotations. also there's the numpy style aligned ones. maybe there are others? inaction is also valid, the more important part is that there are docs rather than their particular format | 06:21:40 |
| K900 | Sphinx is fine | 06:27:47 |
| K900 | PEP727 was never accepted because people tried it and quickly realized it was completely deranged | 06:28:28 |