| 13 Jan 2025 |
 ruro | I am honestly not too familiar with the internals of nixpkgs-review and other CI/automation tooling. The nixpkgs-review README states that it uses ofborg evaluation results do determine, what needs to be built. I wonder if release-cuda.nix could be included in ofborg (and consequently nixpkgs-review) without making hydra.nixos.org build it& | 16:03:37 |
| ruro | * I am honestly not too familiar with the internals of nixpkgs-review and other CI/automation tooling. The nixpkgs-review README states that it uses ofborg evaluation results do determine, what needs to be built. I wonder if release-cuda.nix could be included in ofborg (and consequently nixpkgs-review) without making hydra.nixos.org build it? | 16:03:40 |
| ruro | * I am honestly not too familiar with the internals of nixpkgs-review and other CI/automation tooling. The nixpkgs-review README states that it uses ofborg evaluation results to determine, what needs to be built. I wonder if release-cuda.nix could be included in ofborg (and consequently nixpkgs-review) without making hydra.nixos.org build it? | 16:05:22 |
| 14 Jan 2025 |
 connor (he/him) | When doing review for CUDA that hasn’t been my experience; eval happens locally, consumes a bunch of memory, and then builds stuff
On the other hand, I haven’t run Nixpkgs-review for CUDA stuff since I split it out into a separate repository | 04:14:00 |
 Moritz Sanft | Does anyone of you have some spare time to review https://github.com/NixOS/nixpkgs/pull/372320?
It fixes a silly mistake I made in another PR when writing a patch for libnvidia-container, which unfortunately causes libnvidia-container to segfault under certain conditions. Could possibly mark this as security too, as it's memory corruption :/ | 08:14:41 |
| connor (he/him) | I’m assuming you kept the if expression to minimize the diff in the patch? Otherwise I’d recommend removing it entirely since the condition will always be false. | 16:51:02 |
| connor (he/him) | Merged | 16:58:29 |
|  @maxwell325:matrix.org left the room. | 22:41:36 |
| 15 Jan 2025 |
| Moritz Sanft | In reply to@connorbaker:matrix.org
I’m assuming you kept the if expression to minimize the diff in the patch? Otherwise I’d recommend removing it entirely since the condition will always be false. exactly. Minimizes the room for human error here imo. | 07:23:03 |
|  SomeoneSerge (back on matrix) changed their display name from SomeoneSerge (utc+3) to SomeoneSerge. | 19:02:40 |
| 16 Jan 2025 |
| connor (he/him) | if anyone enjoys the process of bumping opencv looks like 4.11 is out now (master is still using 4.9) | 07:51:29 |
 Gaétan Lepage | Has to happen on staging right ? | 07:52:29 |
| connor (he/him) | I think so due to the number of rebuilds | 07:54:12 |
| connor (he/him) | Good news is that 4.10 fixes compilation with CUDA 12.3+ | 07:54:34 |
| 18 Jan 2025 |
| connor (he/him) | just updated https://github.com/connorbaker/nix-cuda-test to use the latest changes in https://github.com/ConnorBaker/cuda-packages, which include getting rid of the need for a custom CUDA stdenv | 08:51:19 |
| connor (he/him) | At least, I’m fairly certain I’ve implemented it that way. I’ve also got a setup hook which checks for RPATHs in produced libraries that match library directories from NVCC’s host compiler | 09:35:02 |
| Gaétan Lepage | In reply to @connorbaker:matrix.org
if anyone enjoys the process of bumping opencv looks like 4.11 is out now (master is still using 4.9) I'm looking for some reviewers ;)
https://github.com/NixOS/nixpkgs/pull/374246 | 09:45:39 |
| connor (he/him) | I’ll try to take a look this weekend to see if I can help with the tests/fill the missing hashes | 09:53:14 |
 hexa | connor (he/him) (UTC-7): can you respond to the commonts on the onnxruntime PR? https://github.com/NixOS/nixpkgs/pull/364362 | 14:49:05 |
| hexa | I think it is only blocked on those, and I'd like to build upon the PR | 14:49:22 |
| connor (he/him) | How are there so many rebuilds for that PR? Is it taking the number of rebuilds against master? | 18:38:30 |
| hexa | eigen | 18:49:16 |
| hexa | * that is due to the eigen bump | 18:49:22 |
| hexa | * that is due to the eigen bump, I believe | 18:52:05 |
| Gaétan Lepage | On which infra are the cuda derivations (which since a few months, are cached on the nix-community cachix) built ? | 22:57:27 |
| 19 Jan 2025 |
|  Sword joined the room. | 04:21:18 |
| hexa | https://hydra.nix-community.org/project/nixpkgs | 16:10:02 |
| connor (he/him) | Interesting byproduct of getting rid of the CUDA stdenv is having libraries produced by multiple versions of a compiler, which breaks LTO: https://github.com/ConnorBaker/cuda-packages/commit/87cb387ecfba3f46b76e604b896d64417fdf9e8b | 16:15:00 |
| Gaétan Lepage | In reply to @hexa:lossy.network
https://hydra.nix-community.org/project/nixpkgs So on the nix-community CI builders ? | 17:01:35 |
| hexa | yes | 17:03:53 |
