Simon Burge said;

> I can't recall any recent changes to the cache code, and I'm also not
> sure why -current works any better...

My naive suspicion is the insn(s) which makes the error is now populated
in different vaddr than used to be.

Toru Nishimura/ALKYL Technology