| 14 Sep 2025 |
 Alyssa Ross | Not really | 16:59:13 |
| Alyssa Ross | I mean it depends on the BSD | 16:59:25 |
| Alyssa Ross | For FreeBSD the only thing needed was changing some availability macros and getting the event loop to work, either with libepoll-shim or by porting it to kqueue, and that's been a solved problem for years now | 17:00:02 |
| Alyssa Ross | For OpenBSD and NetBSD IIRC Wayland itself works fine but they have a different input stack (wscons) that needed to be supported somewhere. I don't remember if that was in libwayland or libinput or whatever. | 17:00:31 |
| Alyssa Ross | Upstream Wayland supports at least FreeBSD and I think OpenBSD now | 17:00:48 |
| Alyssa Ross | NetBSD was only blocked on having CI last time I looked into it a few years ago | 17:01:13 |
 Randy Eckenrode | That’s better than I thought. Could Wayland on macOS be practical once Mesa’s Vulkan driver lands? | 17:06:22 |
 emily | you'd need a compositor | 17:10:31 |
| emily | and most Wayland stuff will expect to be able to do EGL | 17:10:39 |
| emily | so you'd need EGL | 17:10:44 |
| emily | https://github.com/owl-compositor/owl exists but it doesn't implement any GL stuff so it's kind of useless | 17:10:58 |
 kfiz | kitty just crashed on me with presumably a code signing issue:
-------------------------------------
Translated Report (Full Report Below)
-------------------------------------
Incident Identifier: 53E208A4-1092-48F7-9E5F-DB3B937EA120
CrashReporter Key: 4DCE7E1E-73C0-B340-BF17-C51835E349F9
Hardware Model: MacBookPro18,3
Process: kitty [23039]
Path: /Users/USER/*/kitty.app/Contents/MacOS/kitty
Identifier: net.kovidgoyal.kitty
Version: 0.42.2 (0.42.2)
Code Type: ARM-64 (Native)
Role: Default
Parent Process: launchd [1]
Coalition: net.kovidgoyal.kitty [48843]
Date/Time: 2025-09-14 19:04:14.9866 +0200
Launch Time: 2025-09-14 19:04:14.9744 +0200
OS Version: macOS 15.6.1 (24G90)
Release Type: User
Report Version: 104
Exception Type: EXC_CRASH (SIGKILL (Code Signature Invalid))
Exception Codes: 0x0000000000000000, 0x0000000000000000
Termination Reason: CODESIGNING 4 Launch Constraint Violation
| 17:12:13 |
| Randy Eckenrode | EGL via Zink? | 17:13:16 |
| emily | something like that. | 17:13:37 |
| emily | maybe the gitFull curse | 17:13:51 |
| kfiz | Missed that one. Mind giving me some context? | 17:17:34 |
| emily | https://github.com/NixOS/nixpkgs/issues/208951 – you won't find much insight there though, a lot of people going down blind alleys | 17:18:27 |
| emily | the closest insight we have is Zhaofeng Li's | 17:18:51 |
| emily | which is that it is nondeterministic and depends on whether you have built the derivation on the build system before | 17:19:03 |
| emily | https://github.com/NixOS/nixpkgs/issues/208951#issuecomment-3208584142 has some links to Matrix discussions | 17:19:14 |
| emily | it could be something else entirely too | 17:19:56 |
| emily | but that's the haunted codesigning bug | 17:19:59 |
 K900 | Wayland in practice relies on dma-buf | 17:39:11 |
| K900 | Which BSDs implement because they yoinked all of DRM pretty much | 17:39:20 |
| K900 | owl has a custom protocol for OSX specific buffer passing which is not practical | 17:39:33 |
| emily | you could probably just implement dma-buf. if you really wanted to. | 17:41:28 |
| K900 | Without kernel help? Doubt | 18:17:07 |
|  Vincenzo left the room. | 18:37:24 |
| Vincenzo joined the room. | 18:38:30 |
 n8henrie | Out of general curiosity I decided to enable nixpkgs.config.contentAddressedByDefault on my macbook today. I can't build apple-sdk-11.3:
> Removing the following dylibs from the libSystem reexported libraries list: /usr/lib/system/libcorecrypto.dylib
> /nix/store/accd693hrbxmm122p93v64g9jpmiyj6h-bootstrap-stage0-stdenv-darwin/setup: line 2070: 55582 Killed: 9 llvm-readtapi --filetype=tbd-v5 usr/lib/$libSystem
> 55583 | jq --argjson libs '["/usr/lib/system/libcorecrypto.dylib"]' '
> if .libraries then
> .libraries[] |= select(.install_names[] | any([.] | inside($libs)) | not)
> else
> .
> end
> | .main_library.reexported_libraries[].names[] |= select([.] | inside($libs) | not)
> ' > usr/lib/$libSystem~
For full logs, run:
nix log /nix/store/kslyz4f3vny3bbbddab5f2bwlshccy3l-apple-sdk-11.3.drv
Seeing some out-of-store paths, I tried disabling sandbox, which didn't help.
Is the content-addressed stuff not expected to work (yet) on darwin?
| 18:49:08 |
