Am Mittwoch, 13. Juni 2007 05:02 schrieb chromatic:
> PMC * const key = VTABLE_nextkey_keyed(interp, key_new(interp),
> ns, ITERATE_FROM_START);
>
> If a GC run happens in the rest of this function, it could collect the
> newly created Key -- PMCs created and held outside of structures visible to
> GC are vulnerable to collection.
This really shouldn't happen. All GC runs that are done due to lack of some
ressource are walking the C-stack and the CPU registers I hope[1]. GC runs
initiated at the run-loop level don't need this BTW.
[1] the real problem might be that on MS the marking of the CPU-registers
isn't happening.
See also trace_system_areas() in src/cpu_dep.c and e.g. ...
# ifdef PARROT_HAS_HEADER_SETJMP
Parrot_jump_buff env;
/* Zero the Parrot_jump_buff, otherwise you will trace stale objects */
memset(&env, 0, sizeof (env));
/* this should put registers in env, which then get marked in
* trace_system_stack below
*/
setjmp(env);
# endif
... how to use a standard C lib function to force CPU registers onto the
stack. Other methods for other CPUs are present in that file too.
leo
