| 23 Oct 2025 |
alexfmpe | pkgsCross.ghcjs.haskell.packages.ghc912 has cached miso and reflex-dom, which means a lot of other packages are also cached by being their deps | 17:23:38 |
alexfmpe | I get nix-build -A pkgsCross.ghcjs.haskell.packages.ghc912.ghcjs-base from cache on https://github.com/NixOS/nixpkgs/commit/c90b7f75f4e9d4dffcdc30261f565da1e261d468 which is just a few days old | 17:26:32 |
alexfmpe | magic_rb: ^ | 17:26:42 |
Uraraka ~ Ochaco | I kind of gave up with ghcjs when it started giving cryptic errors so I moved my stack over to wasm | 17:27:18 |
alexfmpe | js backend is recent enough I wouldn't bother much with anything older than 9.12 | 17:27:21 |
alexfmpe | but yeah, wasm is very much an option | 17:27:59 |
alexfmpe | funnily enough eval of this is broken on master and on present haskell-updates, looks like something got merged very recently | 17:28:51 |
alexfmpe | $ nix-build -A pkgsCross.ghcjs.haskell.packages.ghc912.hello
error: no C compiler provided for this platform
| 17:30:22 |
alexfmpe | yay | 17:30:26 |
alexfmpe | bisect was pretty quick: this broke eval: https://github.com/NixOS/nixpkgs/pull/366593 | 17:39:01 |
alexfmpe | somehow we're getting a ghcjs -> emscripten -> openjdk -> python -> openssl that ends up trying to eval stdenv.cc.isGnu that then throws because there's no c compiler | 17:51:32 |
magic_rb | In reply to @alexfmpe:matrix.org js backend is recent enough I wouldn't bother much with anything older than 9.12 mhm i actully figured it out, but i had to apply https://gitlab.haskell.org/ghc/ghc/-/merge_requests/14882 but using sed | 17:54:57 |
magic_rb | had to add both HEAP8 and HEAPU8 | 17:55:12 |
alexfmpe | ah ok, keep in mind I 'backported' that last week | 17:58:28 |
alexfmpe | https://github.com/NixOS/nixpkgs/pull/451527 | 17:58:28 |
magic_rb | ah, good to know | 18:47:06 |
magic_rb | i need to bump nixpkgs anyway due to a broken aeson i think | 18:47:32 |