I tried to address this in the 1.18.0 upstream release, and also added some extra output to the failing test. Unfortunately, it's not enough to pinpoint the problem. I also haven't been able to trigger it on qemu or any of the hardware I have access to.
Is there a way I can set up a test repository to throw a couple of ideas at the wall and try to build it on the failing platform without affecting any of the official repos? -- Hans Petter
