| 16 Nov 2025 |
 hexa | since they are mostly rebuilding the home-assistant tests | 19:57:34 |
| hexa | for which I still have no better idea than exposing them as a top-level package, sorry | 19:57:48 |
| hexa | if they become too annoying we can probably drop the from hydra | 19:58:48 |
| hexa | but it is kinda important to have them in nixpkgs-review results | 19:59:01 |
| 17 Nov 2025 |
 mdaniels5757 | Circling back to this: I was able to build this (/nix/store/gs8wmqs9myb336m5z95prbrn26z9vqdh-firefox-unwrapped-145.0 on aarch64-build-box.nix-community.org), but realized I have no way to test (X11 forwarding doesn't seem to be working, and my virtualization plan has failed on me :)). If someone else happens to have an aarch64-linux machine to test on, that probably would be a good idea. (If you have an aarch64-linux machine, but don't have access to the community builder, I can magic-wormhole you a file that you can use with nix copy --from .) | 01:03:47 |
 dotlambda | Is it acceptable to merge a PR causing <8000 (mostly Python) packages to be rebuilt into staging-next?
https://github.com/NixOS/nixpkgs/pull/462435 | 01:55:06 |
| dotlambda | It's also basically guaranteed to cause no regressions | 01:55:34 |
| dotlambda | The reason for merging into staging-next is that I'm not sure we'll get https://github.com/NixOS/nixpkgs/pull/455889 in before branch-off | 01:59:26 |
| dotlambda | But I'm happy to backport to staging-25.11 instead if that's preferred | 01:59:54 |
| dotlambda | Actually, python313Packages.httpcore (on staging-next) hasn't even been cached by Hydra yet, so this should definitely be fine | 02:06:58 |
| dotlambda | Though python312Packages.httpcore was cached | 02:07:42 |
| dotlambda | In reply to @hexa:lossy.network
if they become too annoying we can probably drop the from hydra I don't see much of a disadvantage in setting hydraPlatforms = [ ] | 02:22:12 |
| dotlambda | Or would that remove them from nixpkgs-review too? | 02:22:34 |
| dotlambda | It would :(
Tested in https://github.com/NixOS/nixpkgs/pull/462445 | 02:35:01 |
|  debtquity joined the room. | 02:54:57 |
 Ihar Hrachyshka | In reply to @yuka:yuka.dev
Is there any reason to assume LLVM 21 miscompilations for aarch64 are limited to aarch64-darwin? I asked in arm channel for someone to test of Firefox crashes on Linux arm too since I don't have a desktop machine myself. There was no reply.
Looking at the patch in llvm I do not think it's specific to Darwin but I can't prove.
And if course Firefox is the app that we know is broken but it's maybe / probably not the only one (though probably also relatively rare). | 02:56:22 |
| Ihar Hrachyshka | * I asked in arm channel for someone to test if Firefox crashes on Linux arm too since I don't have a desktop machine myself. There was no reply.
Looking at the patch in llvm I do not think it's specific to Darwin but I can't prove.
And if course Firefox is the app that we know is broken but it's maybe / probably not the only one (though probably also relatively rare).
| 02:58:14 |
| Ihar Hrachyshka | * I asked in arm channel for someone to test if Firefox crashes on Linux arm too since I don't have a desktop machine myself. There was no reply.
Looking at the patch in llvm I do not think it's specific to Darwin but I can't prove.
And of course Firefox is the app that we know is broken but it's maybe / probably not the only one (though probably also relatively rare). | 02:58:35 |
| Ihar Hrachyshka | * I [asked](https://matrix.to/#/!PXjKZXvVwGOPKtgKti:matrix.org/$Gdbq6AE8KtIhT6FmMCEl2lyhJ1eXjk-z6cac4d6aNTs?via=matrix.org&via=catgirl.cloud&via=nixos.dev) in arm channel for someone to test if Firefox crashes on Linux arm too since I don't have a desktop machine myself. There was no reply.
Looking at the patch in llvm I do not think it's specific to Darwin but I can't prove.
And of course Firefox is the app that we know is broken but it's maybe / probably not the only one (though probably also relatively rare).
| 03:02:18 |
|  adeci joined the room. | 04:26:48 |
 K900 | @Yureka (she/her) so are we sure the Firefox crashes are Mesa and not the LLVM miscompilations? | 07:42:38 |
 Vladimír Čunát | Yes, most of the combinations are in cache.nixos.org at this point. | 07:47:43 |
| Vladimír Čunát | (I looked at 2 pythons * 4 platforms) | 07:48:02 |
| Vladimír Čunát | * (I looked at 2 pythons * 4 platforms via nix-env -qas) | 07:48:10 |
| Vladimír Čunát | And it doesn't seem urgent. | 07:49:20 |
| Vladimír Čunát | Sounds suitable for backport, i.e. just delayed a bit. | 07:49:59 |
| Vladimír Čunát | I switched the PR according to that. | 07:50:30 |
| Vladimír Čunát | * (6/8; I looked at 2 pythons * 4 platforms via nix-env -qas) | 07:52:35 |
 Yureka (she/her) | Yes | 08:18:55 |
| Yureka (she/her) | Because I reproduced a crash with llvm 19 firefoz | 08:19:11 |
