Quite a few of you will have received an email from buildbot reporting apparent failures in 'ninja check-all' on the new clang-cmake-mips builder. This is a configuration mistake on my part and you can safely ignore it.
I'm keeping an eye on the builder and I'm expecting the spurious failures to be resolved in the next build or two.
Sorry for the noise.
The misconfiguration is fixed. I'll give the buildmaster some time to pick up the configuration correction, then force a build from scratch to be sure it's correct. Is there a better way to tell whether it has picked up changes to zorg?
There are two legitimate failures at the moment. Both of them seem to stem from our use of the signext attribute on i32's. I'll fix these by improving the tests.
Both genuine failures have been fixed. This builder is now green.