[gclist] Conservative stack scanning in general
Fri, 27 Jul 2001 07:57:30 -0500
Few programs have trouble with conservative stack scanning and when it
occurs the problem is what I call holes in the stack. These come from
functions with unitialized stack variables. A stack frame with pointers is
poped and a new frame with unitialized variables is pushed in its place.
The collector sees pointers it shouldn't.
This is the worst problem with conservative collection in general, not aunt
Edna's social security number looking like a pointer. The problem is worse
on some machines than others.
A semi cooperating compiler might zero stack pointers after their last use.