| 15 Nov 2025 |
 K900 | Github is being a shit | 15:25:54 |
 hexa | yeah, could only dequeue it from the mq view | 15:25:58 |
| hexa | not from the pr view | 15:26:03 |
| K900 | ffs | 15:26:03 |
 winston | switched the branch :) | 15:28:13 |
| K900 | Requeued | 15:29:58 |
 Ihar Hrachyshka | perl is broken for darwin in staging: https://github.com/NixOS/nixpkgs/pull/462080 | 23:29:58 |
| 16 Nov 2025 |
 Winter | pretty bad darwin regression on master right now, but requires a stdenv rebuild :/ https://github.com/NixOS/nixpkgs/pull/462090 | 02:45:11 |
| Winter | haven’t looked at how far along we are right now but for some reason i doubt we can handle a world rebuild for darwin .-. | 02:49:50 |
 Randy Eckenrode | If we get to the point of doing a world rebuild that includes Linux, I found a regression in curl that breaks Lix for me. It’s pretty niche though. You need to be using an IPv6 link-local DNS server with no other fallback servers. | 06:38:28 |
 Grimmauld (any/all) | an llvm build of modemmanager just sigbus-ed my clang, i hope this was just a flake | 08:45:07 |
| Grimmauld (any/all) | (on current staging-next) | 08:45:25 |
| Grimmauld (any/all) | * (on current staging-next, llvm musl x86_64 linux) | 08:45:51 |
 Vladimír Čunát | We've been running for ~48 hours, so that would be roughly the setback when starting from scratch. (it won't significantly help the real time if it's darwin-only) | 08:52:00 |
| Vladimír Čunát | I'd say this mainly depends on whether the first 25.11 stable would be OK with that regression or not. | 08:54:46 |
| Vladimír Čunát | There's some libxml2 security fix as well, in case we start building from scratch. (and maybe more) | 08:56:31 |
| Grimmauld (any/all) | FAILED: [code=1] src/gallium/drivers/rocket/libintercept.so.p/intercept.c.o
x86_64-unknown-linux-musl-clang -Isrc/gallium/drivers/rocket/libintercept.so.p -Isrc/gallium/drivers/rocket -I../src/gallium/drivers/rocket -Iinclude -I../include -I/nix/store/375gkn4faf8wchh>
../src/gallium/drivers/rocket/intercept.c:299:1: error: conflicting types for 'ioctl'
299 | ioctl(int fd, unsigned long request, ...)
| ^
/nix/store/incvp6dvjg46nsy55y5d32881pqz1cz5-musl-x86_64-unknown-linux-musl-1.2.5-dev/include/sys/ioctl.h:115:5: note: previous declaration is here
115 | int ioctl (int, int, ...);
| ^
1 error generated.
uh oh, musl mesa is acting up now | 10:48:51 |
| K900 | Could try disabling rocket for now | 10:50:49 |
| K900 | You probably don't need it | 10:50:53 |
| Grimmauld (any/all) | yeah thats a fair assumption | 10:55:51 |
| Grimmauld (any/all) | woohoo, toy VM is running on current staging-next, and apart from that mesa thing i got no regressions | 12:01:54 |
 leona | my desktop also runs on staging-next | 12:02:31 |
| Randy Eckenrode | The curl one? It’s caused by a bug in c-ares that was fixed upstream but hasn’t yet made it into a release. I’d be surprised if it’s very common. I wouldn’t restart staging-next just for that, but if it were getting restarted, I think it’s worth including. If it doesn’t make it, I plan on tagging it for backport to 25.11 to be included in the next cycle. | 14:21:21 |
| K900 | Yeah I don't think we restart | 14:51:58 |
| K900 | Timing is already tight | 14:52:01 |
 Yureka (she/her) | Is there any reason to assume LLVM 21 miscompilations for aarch64 are limited to aarch64-darwin? | 17:28:36 |
| Yureka (she/her) | e.g. https://github.com/NixOS/nixpkgs/issues/453372, https://github.com/llvm/llvm-project/issues/166870 | 17:29:19 |
| Yureka (she/her) | miscompilations on both aarch64-{linux,darwin} would be a pretty good reason to scrap this cycle | 17:30:00 |
| hexa |
21k rebuilds on trunk-combined 🫠
| 19:38:35 |
| hexa | * > 21k rebuilds on trunk-combined 🫠 | 19:38:38 |
