On Wed, Apr 10, 2002 at 03:52:09PM -0400, Dan Sugalski wrote: > Can anyone think of a good enough reason to allow unanchored GCable > memory? I can see it being useful for really temporary allocations > that later get dropped, but that seems horribly error-prone, and I'm > not sure I want to set us up for tracking down that sort of thing for > the rest of eternity. > --
You know those places where the documentation says "if you call X() without doing Y, the behavior is undefined"? Well, make X() return a pointer to unanchored GCable memory in those cases. :-)