| 3 Feb 2025 |
 alexfmpe | TL;DR: intentional version warnings are nice to have, but if we can't always guarantee them, then we want the full sweep safety net anyway, in which case no need to clutter hydra logs? | 12:07:38 |
| alexfmpe | * taking dontCheck as an example:
I have a .nix file that filters haskellPackages and filters for all derivations with doCheck = false, then does doCheck = true for all of them | 12:08:00 |
| alexfmpe | * nix-build --keep-going will spit out all the ones that failed - i.e. the ones who need the tests to stay disabled | 12:08:35 |
| alexfmpe | * taking
dontCheck as an example:
I have a .nix file that filters haskellPackages and filters for all derivations with doCheck == false, then does doCheck = true for all of them | 13:51:44 |
| alexfmpe | * taking
`dontCheck` as an example:
I have a .nix file that filters haskellPackages and filters for all derivations with `doCheck == false`, then does `doCheck = true` for all of them | 13:52:03 |
| alexfmpe | * taking
dontCheck as an example:
I have a .nix file that filters haskellPackages for all derivations with doCheck == false, then does doCheck = true for all of them | 13:52:40 |
| alexfmpe | * taking
dontCheck as an example:
I have a .nix file that filters haskellPackages for derivations with doCheck == false, then does doCheck = true for all of them | 13:52:53 |
| alexfmpe | if that step turns out to be wonky, can always inject an easily grepable "I built" on the checked packages
overrideCabal (old: {
postFixup = ''
${old.postFixup or ""}
echo "Override no longer needed for ${old.pname}"
'';
})
| 14:08:36 |
| alexfmpe | $ nix-build -A haskell.packages.ghc98.hello --check 2>&1 | tail -n 3
patching script interpreter paths in /nix/store/6da7xi7dd6wjpgl3bn09xp4p4qjw1ckf-hello-1.0.0.2-doc
Override no longer needed for hello
/nix/store/ais54949224v9ndhgqgj5fz47ah95pij-hello-1.0.0.2
| 14:10:54 |
| alexfmpe | at the cost of extra builds due to --check | 14:11:25 |
| alexfmpe | ah huh actually that won't work because --check errors if the input hasn't been built, would need some force rebuild option | 14:13:03 |
| alexfmpe | annoying that nix doesn't output the built store paths when there were also failures | 14:13:31 |
 Alex | In reply to @alexfmpe:matrix.org
annoying that nix doesn't output the built store paths when there were also failures You could achieve this by splitting your nix-build command into a nix-instantiate (gives drv), a nix-build (realises drv into outputs), and a nix-store --query (can list outputs for drv). | 21:49:16 |
| 4 Feb 2025 |
| alexfmpe | Thanks, I'll try that | 07:19:39 |
|  Axman6 joined the room. | 07:56:52 |
|  unebnf joined the room. | 07:58:14 |
 magic_rb | @axman6:matrix.org youll have to go through pkgsCross.something-something.hostPackages.ghc | 08:06:36 |
| magic_rb | Thats the general way to get a cross compiler with nixpkgs. The javascript one works, havent tested the wasm one | 08:07:06 |
| Axman6 | so pkgsCross.wasi32.hostPackages.ghc121 or something? | 08:07:59 |
| magic_rb | Something along those lines, open up a nix repl and see whats available | 08:08:20 |
| Axman6 | s/Packages/Platform | 08:08:57 |
| Axman6 | yeah doing that now | 08:09:02 |
| magic_rb | In reply to @axman6:matrix.org
s/Packages/Platform Good catch | 08:09:12 |
| magic_rb | You might end up compiling a lot from source, so get a coffee or smth | 08:09:31 |
| Axman6 | hmm, no ghc in there, nor haskell | 08:09:48 |
| magic_rb | Check the JS one, im sure i got that working in the past. So that we confirm were looking in the right place | 08:10:21 |
| Axman6 | .hostPlatform is mostly just info about the host platform, I can't see packages in there (not that I've looked hard) | 08:11:32 |
| magic_rb | Ah then hostPackages? | 08:11:43 |
| magic_rb | Dunno | 08:11:46 |
| Axman6 | there's .gcc but I'd expect that | 08:11:48 |
