| 28 Dec 2024 |
 tuxiqae | In reply to @k900:0upti.me
Yes, that's normal It is? Aren't the linux packages always aligned? | 19:55:19 |
 Atemu | This is the headers | 20:01:34 |
| Atemu | The public API of the kernel doesn't change very often | 20:01:44 |
| Atemu | Certainly not with minor releases | 20:01:55 |
| Atemu | And if you land a patch that does, you're going to get yelled at by Linus | 20:02:41 |
| Atemu | https://lkml.org/lkml/2012/12/23/75 | 20:03:42 |
 SigmaSquadron | The last swear being censored while everything else is not is super funny to me | 20:06:04 |
| tuxiqae | In reply to @atemu12:matrix.org
The public API of the kernel doesn't change very often I just remember that when I used to use Arch (🤢) the Linux and headers were always aligned | 20:18:17 |
 K900 | Well Arch doesn't need a world rebuild to update the kernel headers | 20:21:52 |
| Atemu | Well, they should | 20:22:25 |
| Atemu | It's easy to forget how easy it is to nail down an issue to one particular change in Nix | 20:22:59 |
| Atemu | But we could do the same; it'd just take a staging cycle | 20:23:13 |
| Atemu | But it's just simply not necessary to do very often | 20:23:29 |
| tuxiqae | I see, thanks! | 20:41:17 |
| 30 Dec 2024 |
| Atemu | https://www.phoronix.com/news/RadeonSI-ACO-Default-Pre-GFX10 | 12:19:17 |
| Atemu | Might perhaps be possible to build mesa.drivers without depending on LLVM-lib at runtime soon | 12:19:38 |
| tuxiqae | In reply to @atemu12:matrix.org
Might perhaps be possible to build mesa.drivers without depending on LLVM-lib at runtime soon What's the upside of that? | 12:21:49 |
| Atemu | Leaner closure | 12:21:58 |
| tuxiqae | Leans closer | 12:31:05 |
| K900 | In reply to@atemu12:matrix.org
Might perhaps be possible to build mesa.drivers without depending on LLVM-lib at runtime soon It is already possible, technically | 13:03:04 |
 aktaboot | that would circumvent the rocm breakage? | 13:09:27 |
| K900 | No | 13:10:48 |
| K900 | Mesa uses mainline LLVM | 13:10:58 |
| K900 | Only ROCm itself uses the weird ROCm branch | 13:11:07 |
| K900 | And really fixing the LLVM bits is probably pretty easy | 13:11:37 |
| K900 | It's the rest of that thing that's not | 13:11:41 |
| Atemu | I don't get why it needs to be an LLVM fork | 13:12:07 |
| Atemu | Why not just do it upstream? | 13:12:12 |
| K900 | Because they want to ship it a year ago | 13:12:46 |
| K900 | And not whenever upstream accepts their changes | 13:12:54 |
