| 2 Apr 2026 |
 Randy Eckenrode | The Swift update sucks, and Swift 6.3 appears to change the bootstrap story again, which sucks. It’s very likely the stage 0 bootstrap will be a binary bootstrap that is used to build the Swift compiler that builds the Swift compiler. I hate it, but I don’t think source-based bootstraps are sustainable (especially on Darwin) for Swift. | 11:47:25 |
| Randy Eckenrode | And now I’ve also got to figure out why dotnet is crashing on macOS 26.4 because I actually need that. | 11:47:57 |
| 3 Apr 2026 |
| Randy Eckenrode |
0x106fa4000 - 0x106fa7fff libicucore.A.dylib (*) <30726749-3eae-34a3-8004-08caddbbc2c3> /Volumes/VOLUME/*/libicucore.A.dylib
0x148224000 - 0x1483b3fff libicuuc.76.1.dylib (*) <c8073da6-cb3a-3656-b6bc-c0af25107042> /Volumes/VOLUME/*/libicuuc.76.1.dylib
0x14b604000 - 0x14d66ffff libicudata.76.1.dylib (*) <c9c97135-6a08-351e-9be3-d9b747d4f7b9> /Volumes/VOLUME/*/libicudata.76.1.dylib
0x1487b0000 - 0x1489effff libicui18n.76.1.dylib (*) <79018412-b022-360d-99b8-50334b392861> /Volumes/VOLUME/*/libicui18n.76.1.dylib
0x106fc0000 - 0x106fcbfff libicuio.76.1.dylib (*) <b3f1416f-503e-3dab-8440-e5d69e117a0a> /Volumes/VOLUME/*/libicuio.76.1.dylib
0x148b2c000 - 0x148daffff libclrjit.dylib (*) <1f05f066-7c94-3b78-a0aa-2139b8630d86> /Volumes/VOLUME/*/libclrjit.dylib
0x18cd33000 - 0x18cd7028f libsystem_kernel.dylib (*) <51565b39-f595-3e96-a217-fef29815057a> /usr/lib/system/libsystem_kernel.dylib
0x18cd71000 - 0x18cd7db3b libsystem_pthread.dylib (*) <e7a73008-0c09-31e3-9dd9-0c61652f0e85> /usr/lib/system/libsystem_pthread.dylib
0x18cc06000 - 0x18cc86ef7 libsystem_c.dylib (*) <66ebd321-6899-3863-ba24-5cfc3076a0cb> /usr/lib/system/libsystem_c.dylib
0x190a76000 - 0x190d4bd33 libicucore.A.dylib (*) <bd826ec4-73d6-3f95-abed-001f02b4a429> /usr/lib/libicucore.A.dylib
0x0 - 0xffffffffffffffff ??? (*) <00000000-0000-0000-0000-000000000000> ???
| 13:02:12 |
| Randy Eckenrode | Something is pulling in the system ICU now. 😬 | 13:02:48 |
| Randy Eckenrode | Foundation links against libxml2, which links against libicucore. Anything that links system frameworks will crash using our darwin.ICU build. Great. | 13:28:36 |
| Randy Eckenrode | * | 13:28:56 |
 hexa | does nix-darwin support rebuild/switch over ssh? | 14:58:41 |
| hexa | like for nixos I have nixos-rebuild --target-host root@example.com --flake .#example | 14:59:05 |
| hexa | Redacted or Malformed Event | 14:59:15 |
 Ihar Hrachyshka | yes it definitely works, I do it all the time. what doesn't work is a launchd managed switch (if you use features like copyApps) | 15:29:00 |
| Ihar Hrachyshka | * yes it definitely works, I do it all the time. what doesn't work is a launchd managed switch (if you use features like copyApps)
I think I may have misunderstood what you asked. If you mean native rebuild tool support for --target-host, probably not. I am doing remote repo delivery and run switch from there.
| 16:02:33 |
| hexa | basically rebuilding and switching a remote host from a local tree | 16:22:08 |
| hexa |
| 16:22:14 |
| hexa | infra has been looking at what macs to buy | 16:22:26 |
| hexa | and we're somewhere in the m4 mini (10C, 32GB, 1TB) range | 16:22:49 |
| hexa | do people here have opinions on that choice? | 16:23:21 |
| hexa | Randy Eckenrode, emily, Ihar Hrachyshka? | 16:23:42 |
| hexa | Redacted or Malformed Event | 16:24:05 |
 samasaur | that seems reasonable | 16:25:09 |
| Randy Eckenrode | That seems reasonable. Is that enough RAM? | 16:25:58 |
| Ihar Hrachyshka | this is pretty much the machine I have and it builds LTO firefox in ~ half an hour (if all cores available) | 16:26:40 |
| hexa | we would only get more RAM if we switched to m4 pro cpus | 16:27:21 |
| hexa | or mac studio machines | 16:27:29 |
| hexa | fwiw, we have around 10k EUR to spend for them | 16:27:47 |
| samasaur | for the rebuild/switch question, darwin-rebuild does not support --target-host. Depending on the configuration used you may be able to run the switch over SSH, though FWIW I do my switches over Screen Sharing because some options won't work in a non-GUI session, and worst of all, they don't fail, they stall | 16:29:10 |
| hexa | that sounds horrible | 16:29:38 |
| hexa | no offense | 16:30:09 |
| Ihar Hrachyshka | apple doesn't hear, so no offense :) | 16:30:37 |
| samasaur | unfortunately I agree with you :( | 16:30:40 |
| hexa | we''ll absolutely end up partitioning it | 16:30:48 |
