13 Aug 2024 |
a-kenji | Yes. Sorry I am on mobile right now. | 19:09:44 |
linj | so "build a bit" means have a powerful computer? | 19:10:41 |
emily | or a lot of patience | 19:10:49 |
linj | * so "build a bit" means having a powerful computer? | 19:10:50 |
emily | if you don't have a ton of CPU cycles to spare, you might want to apply for access to some of the community build boxes | 19:10:59 |
linj | oh, good idea | 19:11:23 |
emily | https://nix-community.org/community-builder/ https://github.com/NixOS/aarch64-build-box | 19:11:26 |
linj | I already have access to the darwin box | 19:11:35 |
emily | we have x86_64-linux , aarch64-linux , aarch64-darwin | 19:11:35 |
emily | you'll be competing with me for free CPU cycles though :) | 19:11:46 |
emily | anyway, I think it is also okay to just do some spot checks and let other people review and crunch builds for wide-scale changes | 19:12:26 |
emily | we never go into any staging cycle knowing exactly what will break | 19:12:37 |
linj | I try to improve composability of the rust builder and Emacs lisp builders. Here is my attempt: https://github.com/NixOS/nixpkgs/pull/334476. Any thoughts? | 21:13:43 |
14 Aug 2024 |
| Frank Geusch changed their display name from Master Fudge to Frank Geusch. | 13:25:41 |
| @musicmatze:beyermatthi.as left the room. | 15:21:53 |
emily | I learned today that Cargo will actually refuse to resolve multiple minor versions of a SemVer-compatible major version within one dependency graph even if it's required to satisfy transitive dependencies | 21:04:51 |
emily | https://doc.rust-lang.org/cargo/reference/resolver.html#semver-compatibility | 21:04:58 |
emily | that's very promising for the hopes of only packaging one minor version per major version in Nixpkgs | 21:05:17 |
emily | because it means that =0.1.2 and >0.1 && <0.1.2 are already really fragile in the presence of any other crate that depends on the same crate | 21:05:47 |
15 Aug 2024 |
ˈt͡sɛːzaɐ̯ | Hm, say, if updating to 1.80 in nixpkgs breaks packages because 1.80 adds some warnings and those packages treat warnings as errors, why not set RUSTFLAGS=--cap-lints=warn on those packages and call it a day?
(I do wonder why it's not set globally, too. -Wno-error is not set globally either. Same reason, I suppose.) | 23:54:32 |
16 Aug 2024 |
K900 | It's not that you shouldn't do that | 05:33:49 |
K900 | It's that we haven't done that for staging because there's literally hundreds of those | 05:34:01 |
K900 | And we can't delay staging forever | 05:34:07 |
emily | I'm pretty sure it's a hard error, not a warning? | 05:39:06 |
emily | for the time crate thing | 05:39:15 |
emily | or is this about the cfg thing | 05:39:25 |