| 27 Jun 2026 |
 K900 | But it's fine if you burn it | 22:06:59 |
 emily | I conditionalized it since I'm having to wait for it to get that far on Darwin to confirm the fix anyway | 22:09:59 |
| emily | ok, fixes the build on Darwin | 23:41:06 |
| 28 Jun 2026 |
 ghpzin | python3Packages.numpy consistently fails 1 test on i686-linux for me on staging-next:
FAILED lib/python3.14/site-packages/numpy/lib/tests/test_polynomial.py::TestPolynomial::test_poly_int_overflow - AssertionError:
Arrays are not almost equal to 7 decimals
Mismatched elements: 2 / 21 (9.52%)
Mismatch at indices:
[14]: 1.2066478037803732e+18 (ACTUAL), 1.2066478037803735e+18 (DESIRED)
[15]: -3.599979517947607e+18 (ACTUAL), -3.5999795179476076e+18 (DESIRED)
Max absolute difference among violations: 512.
Max relative difference among violations: 2.12158013e-16
bisected to:
https://github.com/numpy/numpy/commit/c455b03891a5141c87cfd998786111e57cb49fee
revert fixes it, but not sure what it means | 02:24:29 |
| emily | do I want to know what pulls in Numpy on i686? 🫣 | 02:30:27 |
| ghpzin | on staging znes, on nixos pipewire from services.pipewire.alsa.support32Bit | 02:31:46 |
| ghpzin | on staging zsnes, on nixos pipewire from services.pipewire.alsa.support32Bit | 02:32:17 |
| ghpzin | on staging zsnes (and maybe something else)
on nixos pipewire from services.pipewire.alsa.support32Bit = true | 02:34:07 |
| ghpzin | wine and zsnes (maybe something else)
on nixos pipewire from services.pipewire.alsa.support32Bit = true | 02:44:18 |
 Randy Eckenrode | Mmm. Floating point. I wonder if there’s some 80-bit precision happening somewhere. | 03:26:19 |
| K900 | Meanwhile my overnight rebuild ran into a bunch of Python nonsense | 05:34:52 |
| K900 | And now I have to mark 30 different pnpm versions as permittedInsecure I guess | 05:36:10 |
| K900 | Oh no | 05:48:11 |
| K900 | I just realized | 05:48:14 |
| K900 | Python 3.14 | 05:48:19 |
 Grimmauld (any/all) | did we do 3.14 by default? | 06:29:33 |
| K900 | Yes | 06:31:20 |
| Grimmauld (any/all) | fun | 06:31:29 |
| K900 | And there's A LOT | 06:31:32 |
| K900 | As is tradition | 06:31:38 |
| Grimmauld (any/all) | did mypy explode again? | 06:31:48 |
| K900 | Gonna get some coffee and start bonking things | 06:31:50 |
| K900 | Have not gotten that far yet I think | 06:31:58 |
| K900 | But I'm already seeing some really uhh interesting ones | 06:32:21 |
| K900 | Like https://github.com/horazont/aiosasl | 06:33:16 |
| K900 | Not touched since 2022 | 06:33:20 |
 tom | yup, I'd expect a lot of these (unmaintained + asyncio.get_event_loop raising on 3.14). we already fixed a bunch in https://github.com/NixOS/nixpkgs/issues/475732 | 08:58:30 |
| K900 | Yeah I think I'm just going to drop that shit | 08:58:58 |
| K900 | It's got exactly two revdeps | 08:59:22 |
| K900 | aiosasl -> aioxmpp -> urlwatch | 08:59:26 |
