| 23 May 2024 |
connor (he/him) | Updated https://gist.github.com/ConnorBaker/305b1aebd7ee74a258a616bbbd4dcd7b with a list of packages which use requireFile if that's helpful to anyone | 13:36:53 |
connor (he/him) | Gaétan Lepage: if you get a chance, could you patch python312Packages.qgrid? It needs distutils when building with python3.12 or later:
python3.12-qgrid> ModuleNotFoundError: No module named 'distutils'
| 14:28:27 |
Gaétan Lepage | In reply to @connorbaker:matrix.org
Gaétan Lepage: if you get a chance, could you patch python312Packages.qgrid? It needs distutils when building with python3.12 or later:
python3.12-qgrid> ModuleNotFoundError: No module named 'distutils'
Sure ! Done: https://github.com/NixOS/nixpkgs/pull/314010 | 14:43:58 |
connor (he/him) | Would you mind also patching https://github.com/NixOS/nixpkgs/blob/master/pkgs/development/python-modules/svgelements/default.nix?
Maybe like the API for the tests changed (extra s)?
python3.12-svgelements> =================================== FAILURES ===================================
python3.12-svgelements> ___________ TestElementCubicBezierPoint.test_cubic_bounds_issue_220 ____________
python3.12-svgelements> [gw27] linux -- Python 3.12.3 /nix/store/c7ycrgwv039nqglbif98yggx211sdbcl-python3-3.12.3/bin/python3.12
python3.12-svgelements> self = <test.test_cubic_bezier.TestElementCubicBezierPoint testMethod=test_cubic_bounds_issue_220>
python3.12-svgelements> def test_cubic_bounds_issue_220(self):
python3.12-svgelements> p = Path(transform=Matrix(682.657124793113, 0.000000000003, -0.000000000003, 682.657124793113, 257913.248909660178, -507946.354527872754))
python3.12-svgelements> p += CubicBezier(start=Point(-117.139521365,1480.99923469), control1=Point(-41.342266634,1505.62725567), control2=Point(40.3422666342,1505.62725567), end=Point(116.139521365,1480.99923469))
python3.12-svgelements> bounds = p.bbox()
python3.12-svgelements> > self.assertNotAlmostEquals(bounds[1], bounds[3], delta=100)
python3.12-svgelements> E AttributeError: 'TestElementCubicBezierPoint' object has no attribute 'assertNotAlmostEquals'. Did you mean: 'assertNotAlmostEqual'?
python3.12-svgelements> test/test_cubic_bezier.py:82: AttributeError
python3.12-svgelements> ______________________ TestElementWrite.test_write_group _______________________
python3.12-svgelements> [gw20] linux -- Python 3.12.3 /nix/store/c7ycrgwv039nqglbif98yggx211sdbcl-python3-3.12.3/bin/python3.12
python3.12-svgelements> self = <test.test_write.TestElementWrite testMethod=test_write_group>
python3.12-svgelements> def test_write_group(self):
python3.12-svgelements> g = Group()
python3.12-svgelements> > self.assertEquals(g.string_xml(), "<g />")
python3.12-svgelements> E AttributeError: 'TestElementWrite' object has no attribute 'assertEquals'. Did you mean: 'assertEqual'?
python3.12-svgelements> test/test_write.py:27: AttributeError
| 14:49:59 |
connor (he/him) | Oh there's a patch for it! https://github.com/meerk40t/svgelements/commit/23da98941a94cf1afed39c10750222ccfee73c9f | 14:50:30 |
Gaétan Lepage | In reply to @connorbaker:matrix.org Oh there's a patch for it! https://github.com/meerk40t/svgelements/commit/23da98941a94cf1afed39c10750222ccfee73c9f https://github.com/NixOS/nixpkgs/pull/314018 | 14:57:46 |
connor (he/him) | Can you also take a look at https://github.com/NixOS/nixpkgs/blob/master/pkgs/development/python-modules/gpuctypes/default.nix?
When I build it with cudaSupport = true it tries to run tests and fails (note that I've not set testCudaRuntime to true):
python3.12-gpuctypes> ERROR test/test_cuda.py::TestCUDADevice::test_device_count - RecursionError: maximum recursion depth exceeded
python3.12-gpuctypes> ERROR test/test_cuda.py::TestCUDADevice::test_malloc - RecursionError: maximum recursion depth exceeded
| 15:21:03 |
connor (he/him) | Maybe related -- looks like pytest's -k option can't be specified multiple times? I'm seeing behavior where the last value I provide for it is the one it uses | 16:02:45 |
connor (he/him) | I would suggest
pytestFlagsArray =
[ "-v" ]
++ lib.optionals (!testCudaRuntime) [ "--deselect=test/test_cuda.py::TestCUDADevice" ]
++ lib.optionals (!testRocmRuntime) [ "--deselect=test/test_hip.py::TestHIPDevice" ];
| 16:05:19 |
connor (he/him) | (Deselect can be specified multiple times: https://docs.pytest.org/en/latest/example/pythoncollection.html#deselect-tests-during-test-collection) | 16:05:44 |
hexa | hm, can we get rid of these deprecations in 24.05? | 16:23:37 |
hexa | trace: warning: cudaPackages.autoAddDriverRunpath is deprecated, use pkgs.autoAddDriverRunpath instead
trace: warning: cudaPackages.autoAddDriverRunpath is deprecated, use pkgs.autoAddDriverRunpath instead
trace: warning: cudaPackages.autoFixElfFiles is deprecated, use pkgs.autoFixElfFiles instead
trace: warning: cudaPackages.autoAddOpenGLRunpathHook is deprecated, use pkgs.autoAddDriverRunpath instead
trace: warning: cudaPackages.autoAddDriverRunpath is deprecated, use pkgs.autoAddDriverRunpath instead
| 16:23:39 |
hexa | they're not really actionable to the user | 16:24:05 |
connor (he/him) | Where are they coming from? | 16:29:52 |
hexa | good question, let me check my nvidia bits in the config | 18:11:37 |
hexa | { pkgs
, ...
}:
{
nixpkgs.config.cudaSupport = true;
hardware.opengl = {
enable = true;
};
services.xserver.videoDrivers = [
"nvidia"
];
environment.systemPackages = with pkgs; [
cudatoolkit
nvtopPackages.nvidia
];
}
| 18:19:51 |
hexa | so, nothing speical | 18:19:54 |
hexa | tried the two packages, no trace | 18:20:00 |
Gaétan Lepage | In reply to @hexa:lossy.network
trace: warning: cudaPackages.autoAddDriverRunpath is deprecated, use pkgs.autoAddDriverRunpath instead
trace: warning: cudaPackages.autoAddDriverRunpath is deprecated, use pkgs.autoAddDriverRunpath instead
trace: warning: cudaPackages.autoFixElfFiles is deprecated, use pkgs.autoFixElfFiles instead
trace: warning: cudaPackages.autoAddOpenGLRunpathHook is deprecated, use pkgs.autoAddDriverRunpath instead
trace: warning: cudaPackages.autoAddDriverRunpath is deprecated, use pkgs.autoAddDriverRunpath instead
I also have those | 22:07:01 |
hexa | So what is the overlap between our configs? | 22:07:47 |
Gaétan Lepage | I basically have the same config as you have hexa (UTC+1) | 22:18:24 |
Gaétan Lepage | I guess that the origins of those warnings are internal to nixpkgs. | 22:18:40 |
hexa | ok, it is nothing in this file 😄 | 22:41:19 |
hexa | killing nixpkgs.config.cudaSupport gets rid of the trace | 22:41:59 |
hexa | ok, this one is sus | 23:03:45 |
hexa |
trace: warning: cudaPackages.autoAddOpenGLRunpathHook is deprecated, use pkgs.autoAddDriverRunpath instead
| 23:03:51 |
hexa | Already up to date.
❯ rg autoAddOpenGLRunpathHook
nixos/doc/manual/release-notes/rl-2405.section.md
248:- `cudaPackages.autoAddOpenGLRunpathHook` and `cudaPackages.autoAddDriverRunpath` have been deprecated for `pkgs.autoAddDriverRunpath`. Functionality has not changed, but the setuphook has been renamed and moved to the top-level package scope.
pkgs/development/cuda-modules/aliases.nix
15: autoAddOpenGLRunpathHook =
16: mkRenamed "autoAddOpenGLRunpathHook" "pkgs.autoAddDriverRunpath"
| 23:04:40 |
hexa | * ❯ rg autoAddOpenGLRunpathHook
nixos/doc/manual/release-notes/rl-2405.section.md
248:- `cudaPackages.autoAddOpenGLRunpathHook` and `cudaPackages.autoAddDriverRunpath` have been deprecated for `pkgs.autoAddDriverRunpath`. Functionality has not changed, but the setuphook has been renamed and moved to the top-level package scope.
pkgs/development/cuda-modules/aliases.nix
15: autoAddOpenGLRunpathHook =
16: mkRenamed "autoAddOpenGLRunpathHook" "pkgs.autoAddDriverRunpath"
| 23:04:44 |
hexa | the only two mentions | 23:04:49 |
hexa | let's rule out the release notes for a second 😄 | 23:04:55 |