| 22 Dec 2025 |
 Randy Eckenrode | You can mark a package as “edited”. That will cause SwiftPM to use whatever is in Packages/<package-name>. | 02:27:47 |
| Randy Eckenrode | So instead of doing what we were doing, I generate a workspace-state.json that marks all of the dependencies as edited. | 02:28:18 |
| Randy Eckenrode | * So instead of doing what we were doing, I generate a workspace-state.json that marks all of the dependencies as edited; and I symlink the source from the store to Packages/$packageName. | 02:28:57 |
 K900 | Any volunteers to test an mpv update? https://github.com/NixOS/nixpkgs/pull/473235 | 10:10:44 |
 crushing-smite | Is there any foss app launcher that is guaranteed to pick-up darwin/home-manager GUI apps? Don't need nothing fancy, I use sway for NixOS, so I would be happy with just an input field which can filter apps by name. | 13:40:15 |
 kdn | what's some reasonable way to extend arch builder's (or any NixOS VM's) storage with SSD drive on Mac Mini? When I previously created an UTM VM with 500 GB file-backed "disks" they seems to have been corrupted some month in and stopped booting | 13:48:54 |
|  wshtk joined the room. | 14:01:01 |
| wshtk | Hello all. I'm at a loss. Just upgraded to Tahoe the other day... have been running nix-darwin for three or four years on this computer, I rebooted today and suddenly everything was gone, like nix never existed. After hours of troubleshooting, completely reinstalling nix and rebuilding, I'm to the point that it seems that every time I reboot my symlink /run/current-system/ dissapears... rebuild the system or just replace the link, all is fine, reboot /run/current-system gone again. Any ideas what's wiping it? | 14:22:22 |
 Ihar Hrachyshka | Qemu nixpkgs vm hangs on boot sometimes, with traceback in Console that suggests FD_SETSIZE overflow in g_poll. both aarch64 and x86_64 emulated VMs are affected. lsof shows qemu using ~1024+ fds then.
Sometimes the issue happens, sometimes not (so a workaround is rebooting the VM several times until it "settles" on a relatively low number of FDs, ~500-700).
qemu has its own "g_poll" implementation for platforms that support ppoll() that doesn't go through glib loop. AFAIU macos doesn't support it. (?)
is this a known issue? is there anything we could do about it?
| 15:09:23 |
| wshtk | solved: https://github.com/nix-darwin/nix-darwin/issues/1148 | 15:26:20 |
| Randy Eckenrode | As far as I call tell from the man pages, which admittedly kind of suck, all ppoll does is also let you wait on signals. As far as I can tell, g_poll does not let you do that. What is ppoll also doing in this case that one couldn’t do manually with poll? | 15:31:30 |
| Ihar Hrachyshka | qemu g_poll implementation: https://github.com/qemu/qemu/blob/bb7fc1543fa45bebe7eded8115f25441a9fee76e/util/qemu-timer.c#L323-L347 | 15:33:37 |
| Ihar Hrachyshka | in glib, there's this in meson
if host_system in ['windows', 'darwin']
# Poll doesn't work on devices on Windows, and macOS's poll() implementation is known to be broken
glib_conf.set('BROKEN_POLL', true)
endif
| 15:34:28 |
| Randy Eckenrode | I wonder what they mean by “broken”. | 15:35:04 |
| Randy Eckenrode | Sometimes “broken” means “conforms to POSIX but doesn’t do what GNU does”. | 15:35:48 |
| Ihar Hrachyshka | /* The poll() emulation on OS/X doesn't handle fds=NULL, nfds=0,
* so we prefer our own poll emulation.
| 15:35:50 |
| Ihar Hrachyshka | the MR that mentions tap networking broken? https://gitlab.gnome.org/GNOME/glib/-/merge_requests/2571 | 15:39:46 |
| Ihar Hrachyshka | there's a "check" they use to detect a "broken" poll here: https://gitlab.gnome.org/GNOME/glib/-/commit/caecf2dda082e9c46c4157cdc10763deb8dcfc27
but afaiu it is no longer needed and just forced on darwin. wonder if the check would pass now... | 15:40:56 |
| Ihar Hrachyshka | * there's a "check" they use to detect a "broken" poll here: https://gitlab.gnome.org/GNOME/glib/-/commit/caecf2dda082e9c46c4157cdc10763deb8dcfc27
but afaiu it is no longer used and just forced on darwin. wonder if the check would pass now... | 15:41:04 |
| Randy Eckenrode | According to POSIX, fds is an array. My understanding is that NULL is not a valid value for an array in C. | 15:41:55 |
| Ihar Hrachyshka | the check compiled with xcode clang returns 1 | 15:42:22 |
| Randy Eckenrode | https://pubs.opengroup.org/onlinepubs/9799919799/functions/poll.html | 15:42:38 |
| Randy Eckenrode | ppoll is apparently part of POSIX now. No idea if or when Apple will add it. | 15:43:07 |
| Ihar Hrachyshka | they could at least maybe conditionalize it. like if it's null and the user needs some non-standard behavior, go through select. otherwise...
I think main loop for qemu doesn't pass nulls there. | 15:43:45 |
| Ihar Hrachyshka | it's 2024 posix so rather new | 15:43:59 |
| Randy Eckenrode | Yeah. Apple is adding newer stuff, but they still only go for UNIX03 when they certify. | 15:44:53 |
| Randy Eckenrode | So no guarantee, particularly if they have alternative APIs already. | 15:45:24 |
| Randy Eckenrode | Does that means it works now? | 15:46:11 |
| Ihar Hrachyshka | exit(1); /* Does not work for devices -- fail */ | 15:46:34 |
| Randy Eckenrode | That code is almost twenty years old. It wouldn’t be the first time Glib makes an assumption that doesn’t apply on modern Darwin. | 15:46:48 |
