| 30 Jun 2026 |
 matthewcroughan | Anyone getting this when building for musl on staging?
error: Cannot build '/nix/store/y35bzmydcql0sln0c16iww3p6gnlk56v-lkl-2025-11-13.drv'.
Reason: builder failed with exit code 2.
Output paths:
/nix/store/3fbj6ra6lfnbn5g5dbbf9rmjaf441mby-lkl-2025-11-13
/nix/store/3gjknsy2bbmbnd5mbdfys7bzw14dry3k-lkl-2025-11-13-lib
/nix/store/a8zh338klkkbw1ccvvhk7wjm05b47j68-lkl-2025-11-13-dev
Last 25 log lines:
> /build/source/tools/lkl//include/lkl.h:235:52: error: unknown type name ‘mode_t’
> 235 | static inline long lkl_sys_mknod(const char *path, mode_t mode, dev_t dev)
> | ^~~~~~
> In file included from /build/source/tools/lkl//include/lkl_host.h:9,
> from lib/vfio_pci.c:6:
> /build/source/tools/lkl//include/lkl.h:155:52: error: unknown type name ‘mode_t’
> 155 | static inline long lkl_sys_chmod(const char *path, mode_t mode)
> | ^~~~~~
> /build/source/tools/lkl//include/lkl.h:235:65: error: unknown type name ‘dev_t’
> 235 | static inline long lkl_sys_mknod(const char *path, mode_t mode, dev_t dev)
> | ^~~~~
> /build/source/tools/lkl//include/lkl.h:215:52: error: unknown type name ‘mode_t’
> 215 | static inline long lkl_sys_mkdir(const char *path, mode_t mode)
> | ^~~~~~
> /build/source/tools/lkl//include/lkl.h:235:52: error: unknown type name ‘mode_t’
> 235 | static inline long lkl_sys_mknod(const char *path, mode_t mode, dev_t dev)
> | ^~~~~~
> /build/source/tools/lkl//include/lkl.h:235:65: error: unknown type name ‘dev_t’; did you mean ‘div_t’?
> 235 | static inline long lkl_sys_mknod(const char *path, mode_t mode, dev_t dev)
> | ^~~~~
> | div_t
> make[1]: *** [/build/source/tools/build/Makefile.build:106: /build/source/tools/lkl/lib/virtio_net_macvtap.o] Error 1
> make[1]: *** [/build/source/tools/build/Makefile.build:106: /build/source/tools/lkl/lib/vfio_pci.o] Error 1
> make: *** [Makefile:105: /build/source/tools/lkl/lib/liblkl-in.o] Error 2
> LD /build/source/tools/lkl/lklfuse-in.o
For full logs, run:
nix log /nix/store/y35bzmydcql0sln0c16iww3p6gnlk56v-lkl-2025-11-13.drv
| 20:28:02 |
 Sergei Zimmerman (xokdvium) | Ugh, curl thingy did fuck up nix too. We just have more a bounded wakeup on the curl worker thread | 23:12:35 |
| Sergei Zimmerman (xokdvium) | Nix on 26.05 is fucked | 23:13:00 |
| Sergei Zimmerman (xokdvium) | Seems to have fucked it in a hard to notice way though, i.e. most stuff is fine but there are cases when the worker thread is slow to start | 23:30:28 |
 emily | so what are we doing for libssh2? | 23:42:47 |
| emily | RCE with a public PoC seems pretty bad | 23:43:02 |
| emily | if we pick the fix patches, then we can also bump Rust to 1.96.1 for free to pick up the miscompilation fix | 23:43:33 |
| Sergei Zimmerman (xokdvium) | FWIW https://github.com/NixOS/nixpkgs/pull/537230, I don't know whether we want to eat the rebuilds. | 23:48:36 |
 hexa | staging-nixos-26.05 | 23:51:24 |
| hexa | because it will certainly rebuild all nixos tests | 23:51:46 |
| Sergei Zimmerman (xokdvium) | Oh we have it now? | 23:52:45 |
| hexa | we do | 23:52:49 |
| Sergei Zimmerman (xokdvium) | K | 23:52:50 |
| hexa | the bot should yell at you shortly | 23:52:56 |
| hexa | tada.wav | 23:53:10 |
| emily | Sergei Zimmerman (xokdvium): do you also need to revert it on 26.11? | 23:55:33 |
| emily | Lix carries a revert now I think? | 23:55:36 |
| Sergei Zimmerman (xokdvium) | Right yeah, I suppose so. Though the breakage seems rather scoped (i.e. 10s wait under yet undeterminted conditions) | 23:57:46 |
| hexa | then I gues via staging-next or staging-nixos, whichever comes first | 23:58:14 |
| emily | hexa: any thoughts on ^? | 23:59:59 |
| 1 Jul 2026 |
| hexa | pretty sure lesuisse already responded on the closed pr | 00:00:15 |
| hexa | https://github.com/NixOS/nixpkgs/pull/533237#issuecomment-4844060191 | 00:00:30 |
| emily | agreed on picking vs. bumping | 00:00:50 |
| emily | I'm more concerned about whether we put it in staging-next, given the severity | 00:00:59 |
| emily | it's ~50k rebuilds per platform but the latest CVE seems pretty awful | 00:01:21 |
| emily | and iirc wasn't out at the time of that PR? | 00:01:28 |
| hexa | annoying | 00:02:23 |
| emily | it does give us the chance to pick up Rust 1.96.1 if we do it, since that was rushed out to ship both the libssh2 fixes and fix a miscompilation bug, though AFAICT the miscompilation bug isn't bad enough to be worth the rebuilds unless we're eating them anyway because of libssh2 | 00:03:35 |
| emily | we are sufficiently ahead of schedule that it feels like it might be worth eating the rebuilds rather than shipping an RCE for weeks on unstable, but don't feel confident enough to make the call myself. worried about a call not happening until it's throwing away even more builds / delaying things longer though | 00:05:02 |
| hexa | let's just do it | 00:07:31 |
