[RFA] Nullified garbage-collected global variables

[Jerome Guitton]

Tom Tromey (tromey@redhat.com):

> Do you need to apply the same treatment to full symbols?  What happens
> if the psymtab is expanded for some other reason and a full symbol of
> this sort is then created?  Or is that already impossible?

Well, that sounded unlikely; then I double-checked, and experimental
results rejected my assumption. e.g. my example without --readnow:

(gdb) p my_global_symbol
No symbol "my_global_symbol" in current context.

and with --readnow:

(gdb) p my_global_symbol 
Cannot access memory at address 0x0

Thank you for catching that. I'll fix my patch.


> What about has_section_at_zero?

Right, it's worth protecting us against this case as well (as we do for
elminated functions). I'll add a guard.

Thank you for the review,
Jerome