This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 03/11] Iterate over 'struct varobj_item' instead of PyObject
- From: Keith Seitz <keiths at redhat dot com>
- To: Yao Qi <yao at codesourcery dot com>, gdb-patches at sourceware dot org
- Date: Tue, 21 Jan 2014 12:44:23 -0800
- Subject: Re: [PATCH 03/11] Iterate over 'struct varobj_item' instead of PyObject
- Authentication-results: sourceware.org; auth=none
- References: <1385258996-26047-1-git-send-email-yao at codesourcery dot com> <1385258996-26047-4-git-send-email-yao at codesourcery dot com>
On 11/23/2013 06:09 PM, Yao Qi wrote:
gdb:
2013-11-24 Pedro Alves <pedro@codesourcery.com>
Yao Qi <yao@codesourcery.com>
* python/py-varobj.c (py_varobj_iter_next): Move some code
from varobj.c.
(py_varobj_get_iterator):
^^^^^^^^^^^^^^^^^^^^^^
Superfluous? I don't see any changes to that function.
* varobj-iter.h (struct varobj_item): New. Moved from
varobj.c.
I would remove "New." That threw me a little while reading the patch. It
looked familiar! Simply saying that you moved it is sufficient.
* varobj.c: Move "varobj-iter.h" inclusion earlier.
(struct varobj_item): Remove.
I think it is more concise to say "Moved to varobj-iter.h" instead of
"Remove".
(varobj_clear_saved_item): New function.
(update_dynamic_varobj_children): Move python-related code to
py-varobj.c.
(free_variable): Call varobj_clear_saved_item.
---
gdb/python/py-varobj.c | 17 ++++++++++-
gdb/varobj-iter.h | 12 ++++++-
gdb/varobj.c | 73 ++++++++++++++++-------------------------------
3 files changed, 51 insertions(+), 51 deletions(-)
diff --git a/gdb/python/py-varobj.c b/gdb/python/py-varobj.c
index 6588fc2..506e865 100644
--- a/gdb/python/py-varobj.c
+++ b/gdb/python/py-varobj.c
@@ -51,6 +51,9 @@ py_varobj_iter_next (struct varobj_iter *self)
struct py_varobj_iter *t = (struct py_varobj_iter *) self;
struct cleanup *back_to;
PyObject *item;
+ PyObject *py_v;
+ varobj_item *vitem;
+ const char *name = NULL;
back_to = varobj_ensure_python_env (self->var);
@@ -98,9 +101,21 @@ py_varobj_iter_next (struct varobj_iter *self)
}
}
+ if (!PyArg_ParseTuple (item, "sO", &name, &py_v))
+ {
+ gdbpy_print_stack ();
+ error (_("Invalid item from the child list"));
+ }
+
+ vitem = xmalloc (sizeof *vitem);
You can use "XNEW (struct varobj_item)" here, too, although there isn't
any real overriding reason to do so (other than personal preference).
Feel free to ignore me, though. There is no hard rule for this in
gdb-land as far as I know.
+ vitem->value = convert_value_from_python (py_v);
+ if (vitem->value == NULL)
+ gdbpy_print_stack ();
+ vitem->name = xstrdup (name);
+
self->next_raw_index++;
do_cleanups (back_to);
- return item;
+ return vitem;
}
/* The 'vtable' of pretty-printed python varobj iterators. */
diff --git a/gdb/varobj-iter.h b/gdb/varobj-iter.h
index 59be278..b4bda82 100644
--- a/gdb/varobj-iter.h
+++ b/gdb/varobj-iter.h
@@ -14,9 +14,17 @@
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>. */
-struct varobj_iter_ops;
+/* A node or item of varobj, composed by the name and the value. */
+
+typedef struct varobj_item
+{
+ /* Name of this item. */
+ char *name;
+ /* Value of this item. */
+ struct value *value;
+} varobj_item;
[Changes mentioned in last patch for above hunk]
-typedef PyObject varobj_item;
+struct varobj_iter_ops;
/* A dynamic varobj iterator "class". */
diff --git a/gdb/varobj.c b/gdb/varobj.c
index 7a8305b..28e934c 100644
--- a/gdb/varobj.c
+++ b/gdb/varobj.c
@@ -33,6 +33,7 @@
#include "vec.h"
#include "gdbthread.h"
#include "inferior.h"
+#include "varobj-iter.h"
#if HAVE_PYTHON
#include "python/python.h"
@@ -41,8 +42,6 @@
typedef int PyObject;
#endif
-#include "varobj-iter.h"
-
/* Non-zero if we want to see trace of varobj level stuff. */
unsigned int varobjdebug = 0;
@@ -110,16 +109,6 @@ struct varobj_root
struct varobj_root *next;
};
-/* A node or item of varobj, composed by the name and the value. */
-
-struct varobj_item
-{
- /* Name of this item. */
- char *name;
- /* Value of this item. */
- struct value *value;
-};
-
/* Dynamic part of varobj. */
struct varobj_dynamic
@@ -787,6 +776,18 @@ varobj_get_iterator (struct varobj *var)
requested an interator from a non-dynamic varobj"));
}
+/* Release and clear VAR's saved item, if any. */
+
+static void
+varobj_clear_saved_item (struct varobj_dynamic *var)
+{
+ if (var->saved_item != NULL)
+ {
+ value_free (var->saved_item->value);
+ xfree (var->saved_item);
+ var->saved_item = NULL;
+ }
+}
#endif
static int
@@ -801,14 +802,8 @@ update_dynamic_varobj_children (struct varobj *var,
int to)
{
#if HAVE_PYTHON
- struct cleanup *back_to;
int i;
- if (!gdb_python_initialized)
- return 0;
-
- back_to = varobj_ensure_python_env (var);
-
*cchanged = 0;
if (update_children || var->dynamic->child_iter == NULL)
@@ -816,16 +811,12 @@ update_dynamic_varobj_children (struct varobj *var,
varobj_iter_delete (var->dynamic->child_iter);
var->dynamic->child_iter = varobj_get_iterator (var);
- Py_XDECREF (var->dynamic->saved_item);
- var->dynamic->saved_item = NULL;
+ varobj_clear_saved_item (var->dynamic);
i = 0;
if (var->dynamic->child_iter == NULL)
- {
- do_cleanups (back_to);
- return 0;
- }
+ return 0;
}
else
i = VEC_length (varobj_p, var->children);
@@ -834,10 +825,10 @@ update_dynamic_varobj_children (struct varobj *var,
are more children. */
for (; to < 0 || i < to + 1; ++i)
{
- PyObject *item;
+ varobj_item *item;
/* See if there was a leftover from last time. */
- if (var->dynamic->saved_item)
+ if (var->dynamic->saved_item != NULL)
{
item = var->dynamic->saved_item;
var->dynamic->saved_item = NULL;
@@ -845,6 +836,10 @@ update_dynamic_varobj_children (struct varobj *var,
else
{
item = varobj_iter_next (var->dynamic->child_iter);
+ /* Release vitem->value so its lifetime is not bound to the
+ execution of a command. */
+ if (item != NULL && item->value != NULL)
+ release_value_or_incref (item->value);
}
if (item == NULL)
@@ -857,36 +852,19 @@ update_dynamic_varobj_children (struct varobj *var,
/* We don't want to push the extra child on any report list. */
if (to < 0 || i < to)
{
- PyObject *py_v;
- const char *name;
- struct varobj_item varobj_item;
- struct cleanup *inner;
int can_mention = from < 0 || i >= from;
- inner = make_cleanup_py_decref (item);
-
- if (!PyArg_ParseTuple (item, "sO", &name, &py_v))
- {
- gdbpy_print_stack ();
- error (_("Invalid item from the child list"));
- }
-
- varobj_item.value = convert_value_from_python (py_v);
- if (varobj_item.value == NULL)
- gdbpy_print_stack ();
- varobj_item.name = xstrdup (name);
-
install_dynamic_child (var, can_mention ? changed : NULL,
can_mention ? type_changed : NULL,
can_mention ? new : NULL,
can_mention ? unchanged : NULL,
can_mention ? cchanged : NULL, i,
- &varobj_item);
- do_cleanups (inner);
+ item);
+
+ xfree (item);
}
else
{
- Py_XDECREF (var->dynamic->saved_item);
var->dynamic->saved_item = item;
/* We want to truncate the child list just before this
@@ -2180,12 +2158,11 @@ free_variable (struct varobj *var)
Py_XDECREF (var->dynamic->constructor);
Py_XDECREF (var->dynamic->pretty_printer);
- Py_XDECREF (var->dynamic->child_iter);
If I'm reading this right, this statement was moved to the python
iterator dtor function. What happens if we do not iterate over all
children and then delete the varobj? I don't see where
varobj_delete_iter is called in that case, i.e., isn't this leaked in
this scenario?
- Py_XDECREF (var->dynamic->saved_item);
do_cleanups (cleanup);
}
#endif
+ varobj_clear_saved_item (var->dynamic);
value_free (var->value);
/* Free the expression if this is a root variable. */