Using the kernel from the December 9,2012, i386 current snapshot, I am able to successfully build xenocara without incident; however, using a custom kernel option NKPTP=12 that succeeded in building userland successfully, I see the same problems building xenocara as I saw building userland with the GENERIC kernel.
After I installed the December 9 i386 snapshot and tried to update current source on around the night of December 10, I had the impression there had been a regression sometime between the December 9 snapshot and early December 11. The closest candidates appear to be 2012/12/10 15:42:54 sys/uvm : uvm_pdaemon.c 2012/12/10 15:58:04 sys/ufs/ffs : ffs_softdep.c Of course the complicating factor is that the problem has only been observed running current i386 on one machine whereas running current amd64 on the same machine exhibits no similar problems.
