So, with some minor modifications*, none worthy of changelog entries:
I just fixed a fatal triggered by sprintf("%O") on mappings under some circumstances. There also seems to be an issue with the machine-code generation on sparc in combination with gcc 4.3.4.