| 1 Dec 2025 |
 alexfmpe | am I holding it wrong or x86_64 -> aarch32 works while aarch64->aarch32 doesn't ?
$ uname -mo
Darwin arm64
$ nom-build -A pkgsCross.armv7a-android-prebuilt.hello --argstr system "aarch64-linux"
error: Android NDK doesn't support building on aarch64-unknown-linux-gnu, as far as we know
$ nom-build -A pkgsCross.armv7a-android-prebuilt.hello --argstr system "x86_64-linux"
/nix/store/2nnknh2lbi49n97n8cii6lqgr64bfzjp-hello-armv7a-unknown-linux-androideabi-2.12.2
| 18:24:28 |
| alexfmpe | (using rosetta/linux builder shenanigans) | 18:24:54 |
 K900 | On Darwin? Very possibly | 18:25:20 |
| K900 | lol | 18:25:21 |
| K900 | Though even not on Darwin I'm guessing the Android SDK does not have aarch64 host tools | 18:25:46 |
 emily | --argstr system "aarch64-linux" should not be Darwin | 18:26:09 |
| alexfmpe | eval = aarch64-darwin
build = x86_64-linux
host = aarch64-linux | 18:39:22 |
| alexfmpe | I think? | 18:39:24 |
| alexfmpe | * eval = aarch64-darwin
build = aarch64-linux
host = aarch64-linux
eval = aarch64-darwin
build = x86_64-linux
host = aarch64-linux | 18:39:59 |
| alexfmpe | * Those two commands mean
eval = aarch64-darwin
build = aarch64-linux
host = aarch64-linux
eval = aarch64-darwin
build = x86_64-linux
host = aarch64-linux | 18:40:11 |
| K900 | Eval should not matter at all | 18:40:34 |
| K900 | Anyway, the NDK does not in fact have aa64 host tools | 18:40:43 |
| K900 | So yes | 18:40:50 |
| alexfmpe | right, I'm just clarifying where 'darwin' fits in | 18:41:00 |
| emily | no AArch64 surprises me. you can't do Android dev on AArch64 Chromebooks? | 18:52:47 |
| K900 | You can't even run an IDE on those | 18:55:46 |
| K900 | Can you | 18:55:49 |
| emily | sure you can | 20:05:35 |
| emily | they have built-in Linux VMs with full graphical Wayland app support | 20:05:42 |
| emily | Android Studio is explicitly supported on the Intel ones at least | 20:05:50 |
| emily | no idea about the ARM ones but it'd surprise me if not | 20:05:57 |
| 2 Dec 2025 |
 DavHau | Occasionally I encounter cross packages which simply compile binaries for the wrong platform.
Is there a hook in nixpkgs that I can use to check if all binaries in $out/bin are for the correct platform? | 06:43:18 |
| K900 | I don't think so, but it would be nice to have one | 06:49:49 |
| DavHau | done: https://github.com/NixOS/nixpkgs/pull/467183 | 16:04:09 |
| K900 | I wonder how much it'll suck to just add it by default | 16:08:43 |
 hexa | could this work with python3Minimal? | 16:10:25 |
| DavHau | Let's add it by default later. I would like to make it available at all for now.
And yeah, good point, it should use python3Minimal | 16:11:49 |
| hexa | also not sure how fancy we want to go, but couldn't the hook be null in non-cross situations | 16:23:54 |
| hexa | that way we wouldn't have to guard it on every use | 16:24:14 |
| DavHau | I think it would be better to go the route to always include it by default. even in non-cross scenarios. the check is very cheap. | 16:25:21 |
