Martin Husemann <[email protected]> writes: > On Tue, Oct 27, 2026 at 08:27:17PM -0400, Greg Troxel wrote: >> I rolled back the kernel, and am working on libs, but it seems something >> has gone wrong. > > It is libc, and has been fixed on the -10 branch: > > https://mail-index.netbsd.org/source-changes/2026/10/26/msg164819.html
Thanks. This incident to me is a really big deal. For as long as I can remember, going back to 2010ish, updating sources along a stable branch, and then if it built, updating and rebooting, has been 100% reliable. I do updates on my pkg builder domUs first anyway, but as those don't run named I didn't notice. This resulted in over an hour of service downtime. I realize that when people submit pullups they think everything is fine, and it's probably too hard to run anita every time. I don't even know if our atf tests would have caught this. I didn't notice discussion that there was a system-breaking bug in the branch. It would be good to send a note with the Big Scary Subject to netbsd-users when this sort of thing is first discovered, so people who have done a cvs update, and then are in the process of build, sync, update can be warned. As in "netbsd-10 branch is broken; don't update!". Also, it feels to me like the rate of changes that seem kind of big for the branch has increased. That could be because of a higher rate of vulnerability discovery, but overall the stable branch feels less stable. But looking at CHANGES, it seems this particular case was fixing a real problem on other arches, and the pulled-up fix was just incomplete.
