One more patch. Turns out I missed a small detail in my red-black tree swap-out. The RBI_INIT macro was doing double-duty initializing both its own rb-tree pointers and the rest of the fields that happened to be in the same struct. So, the native-rb-tree version wasn't properly initializing the rest of the fields in the struct. Just moving one line fixes that problem. - Geoff
Attachment:
ipf-rb-init.patch
Description: Binary data