FreeChainXenon/binutils-gdb
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix pr 16612.

Browse source 
* guile/scm-type.c (tyscm_copy_type_recursive): Move type to its
	new eq?-hashtab.

	testsuite/
	* gdb.guile/scm-value.ep (test_value_after_death): Do a garbage
	collect after discarding symbols.
This commit is contained in:
Doug Evans 2014-03-13 09:55:12 -07:00
parent 350e1a768c
commit 5a1e8c7a83
4 changed files with 34 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

19
gdb/guile/scm-type.c
View file

@ -363,11 +363,30 @@ tyscm_copy_type_recursive (void **slot, void *info)
type_smob *t_smob = (type_smob *) *slot;
htab_t copied_types = info;
struct objfile *objfile = TYPE_OBJFILE (t_smob->type);
htab_t htab;
eqable_gdb_smob **new_slot;
type_smob t_smob_for_lookup;
gdb_assert (objfile != NULL);
htab_empty (copied_types);
t_smob->type = copy_type_recursive (objfile, t_smob->type, copied_types);
/* The eq?-hashtab that the type lived in is going away.
Add the type to its new eq?-hashtab: Otherwise if/when the type is later
garbage collected we'll assert-fail if the type isn't in the hashtab.
PR 16612.
Types now live in "arch space", and things like "char" that came from
the objfile *could* be considered eq? with the arch "char" type.
However, they weren't before the objfile got deleted, so making them
eq? now is debatable. */
htab = tyscm_type_map (t_smob->type);
t_smob_for_lookup.type = t_smob->type;
new_slot = gdbscm_find_eqable_gsmob_ptr_slot (htab, &t_smob_for_lookup.base);
gdb_assert (*new_slot == NULL);
gdbscm_fill_eqable_gsmob_ptr_slot (new_slot, &t_smob->base);
return 1;
}