| 17 Mar 2026 |
 Artemis Tosini (NixOS) | nixpkgs can do ppc32, which go won't do | 19:00:32 |
 Puna | yeah i guess no Go isn't much of a blocker | 19:15:21 |
| Puna | (it's also a no-go on ppc64 cus they raised their baseline way too high, and refuse to undo it) | 19:17:11 |
|  colemickens joined the room. | 19:36:22 |
|  @panic:sft.lol left the room. | 20:36:06 |
| 18 Mar 2026 |
|  aphasic joined the room. | 06:35:06 |
| Artemis Tosini (NixOS) | the most cursed architecture one could usefully add to nixpkgs might be netbsd-vax. Relies on somewhat better netbsd support though, most of the BSD work has been focused on FreeBSD | 13:34:01 |
| Artemis Tosini (NixOS) | FreeBSD is fairly aggresive about dropping platform support, the only supported architectures for 16 are amd64, aarch64, armv7, powerprc64 (big endian), powerpc64le, and riscv64 (hard float) | 13:35:34 |
| 20 Mar 2026 |
|  irrelevancy changed their display name from irrelevancyisthygoal One Confused Being to irrelevancy. | 06:46:49 |
 alex taxia | In reply to @artemist:nixos.dev
the most cursed architecture one could usefully add to nixpkgs might be netbsd-vax. Relies on somewhat better netbsd support though, most of the BSD work has been focused on FreeBSD thinking about watching the tapes spin while it builds | 22:25:36 |
| 24 Mar 2026 |
 alexfmpe | what counts as exotic again?
tier 3 and above or what? | 03:31:04 |
| alexfmpe | not directly, but I was kicking the tires in haskell cross and ran into a s390x thing: https://github.com/NixOS/nixpkgs/pull/502917 | 03:32:02 |
| Puna | RFC 46 said that the exotic-platform-maintainers github team can be consulted for tier 4, but i'd imagine anything with less support also goes. tier 3 apparently includes loongarch64, the MIPSes, 64-bit POWER, and s390x though, which i would all consider to be some kind of exotic. | 10:27:32 |
 7c6f434c | RFC 46 sets average expectations, surpassing them is welcome! | 10:44:50 |
|  Lotte (it/its)/Cinny (she/her) θΔ& changed their profile picture. | 12:15:07 |
| 27 Mar 2026 |
 Tristan Ross | Yeah, RFC 46 sets what to expect but a good things to look at is the breakdown table in the manual. https://nixos.org/manual/nixpkgs/stable/#sec-platform-breakdown | 15:34:57 |
| 31 Mar 2026 |
 Alyssa Ross | qemu-user officially orphaned upstream cc dramforever | 10:19:43 |
| Alyssa Ross | https://gitlab.com/qemu-project/qemu/-/commit/e813d3be36cf9269ca15764a3de3473149d8a022 | 10:23:06 |
 Stas | i always assumed that the userspace qemu is the more elegant one | 10:28:23 |
| Alyssa Ross | very much not the case | 10:28:51 |
| Alyssa Ross | it has to implement the whole Linux API, so you'll frequently run into stuff that isn't implemented or doesn't work right | 10:29:11 |
| Alyssa Ross | massively bigger surface than emulating hardware | 10:29:19 |
| Stas | true that :( When it runs well, it is like it is not there though | 10:29:38 |
| Stas | I use it to compile all my arm64 builds, not cross compile | 10:29:54 |
| Stas | works perfectly, albeit slower | 10:29:59 |
| Stas | * works perfectly, albeit slower than native of course | 10:30:05 |
| Alyssa Ross | a VM set up as a remote builder is also like it's not there, if you're doing Nix builds, and much more reliable | 10:30:30 |
| Stas | I am considering it, as I want the original 32bit arm too. Hetzner's Neoverse cores do not support the old isa though :( | 10:31:29 |
| Lotte (it/its)/Cinny (she/her) θΔ& | i wonder if such a vm would alleviate the cpu cache thrashing that i have noticed with qemu-user because you can effectively limit and/or isolate the emulation to specific cpu cores | 10:36:46 |
| Lotte (it/its)/Cinny (she/her) θΔ& | * i wonder if such a vm would alleviate the cpu cache thrashing that i have noticed with qemu-user because you can more effectively limit and/or isolate the emulation to specific cpu cores | 10:36:55 |
