| 13 Aug 2025 |
 emily | I think the real question is whether you can build a working rustc using a rustc with the Cranelift backend | 18:55:35 |
| emily | i.e. whether it can bootstrap | 18:55:40 |
| emily | (also maybe someone should… fix the comment?) | 18:55:46 |
 rosssmyth | Apparently, yes. As there is a test in the cranelift backend repo for it | 18:57:59 |
| emily | got a link? | 18:58:44 |
| rosssmyth | https://github.com/rust-lang/rustc_codegen_cranelift/blob/master/scripts/test_bootstrap.sh | 18:58:50 |
| emily | isn't that just checking that you can build rustc once? | 19:02:01 |
| emily | I'm talking about | 19:02:04 |
| emily |
- build Rust n with only the Cranelift backend
- use that Rust to build Rust n+1
| 19:02:15 |
| emily | this only does --stage 1, so I think it's just checking you can build "rustc with Cranelift backend" using an existing Rust binary | 19:02:29 |
| emily | maybe I'm wrong though and the rustc in the environment here is a Cranelift one? | 19:02:41 |
| rosssmyth | Yes, I believe you are correct | 19:03:14 |
 Alyssa Ross | mrustc apparently supports 1.74.0 anyway now which should be a significantly newer LLVM | 19:09:50 |
| emily | only 16 | 19:22:26 |
| emily | which I am also dropping | 19:22:28 |
| emily | 🔪 | 19:22:41 |
| emily | 12, 13, 14, 15, 16, 17 are all dying | 19:22:56 |
| emily | anyway the mrustc-bootstrap compiler is useless | 19:23:33 |
| emily | because it gives you a Rust too old for Nixpkgs | 19:23:36 |
| emily | and no chain from there to the Rust we use | 19:23:42 |
| emily | so I am only marking it broken and giving pointers to how you could resurrect it in a comment | 19:23:53 |
| emily | actually maybe 17 too | 19:24:10 |
| emily | Cranelift or backporting LLVM 18 support patches to 1.74 or hoping mrustc adds support for new versions all seem like viable paths forward if you cared about it | 19:24:21 |
| emily | my suspicion is nobody does | 19:24:24 |
 Tristan Ross | In reply to @emilazy:matrix.org
12, 13, 14, 15, 16, 17 are all dying Excellent, I support that | 19:24:47 |
| emily | GCC 9, 10, 11, 12 too | 19:25:05 |
| Tristan Ross | 18, 19, 20, 21, and git should suffice | 19:25:10 |
| emily | I have every package that depends on any of those either dropped, marked broken, or fixed-in-principle locally | 19:25:14 |
| emily | just a lot of testing and fixups to do | 19:25:19 |
| Tristan Ross | Awesome, that GCC stuff will make it easier to move to GCC NG. | 19:25:41 |
