Re: [GLIB] memory leak in glib 2.32.4 ?



Hi,

Noone to give me a support on this ? It is the bad place to ask this
kind of question ? Do I have to find another mailing list ? This
mailing list seems to be deserted. I also try to use last version of
dbus, glib-dbus, glib to see if I am able to fix my leak but no
improvement.

I also recode from scratch a client and a server which are very
simple. I am able to run it on a fedora 16 without leak. If I use it
on my embedded device... leak is always there. Source can be found
here : http://arthurlambert.fr/robotique/test_dbus.tar.bz2

It is quite sad, I can identify which lines are causing a leak without
be able to find a fix !

Regards,
Arthur

2014-03-06 14:11 GMT+01:00 Arthur Lambert <lambertarthur22 gmail com>:
Hi Simon,

First, when I put all the free that I have to put, the code get a
segfault. In fact It seems that I just have to use g_ptr_array_free on
the upper structure that is the ptr_array to free all memory. If I put
more free, I get plenty of invalid free though valgrind test and
segfault.

Second, How can you explain that I am not able to see any leak with ps
aux command if I put a loop in my main (or only on some specific
target) ? Do they have a way to garbarge memory in glib ?

Last point is I am currently using this code in dbus. I cannot use
free because all this structure is send though dbus. I get a pointer
by argument on a callback. I use the pointer to set the answer. So I
suppose that dbus must free all the structure by itself riight ?

Regards,
Arthur.

2014-03-06 12:58 GMT+01:00 Simon K├ągedal Reimer <skagedal gmail com>:
Hi Arthur,

In C code, memory is managed manually. Every time you dynamically allocate
memory, such as with GLib's various _new() functions, you need to return
that chunk of memory to the system when you are done using it. I see no such
calls in your code. For example, a GPtrArray created with g_ptr_array_new()
should have a matching call to g_ptr_array_unref() or g_ptr_array_free(). A
call to g_new0 should have a matching g_free, I don't know why you have
commented that out. And so on.

Best regards! Simon


On Tue, Mar 4, 2014 at 3:44 PM, Arthur Lambert <lambertarthur22 gmail com>
wrote:


Hi All,

I am using glib with GValue, GPtrArray to communicate with dbus. I
have a big memory leak in my code. I am running an embedded target
with glib 2.32.4. I am currently trying to  code some little binary to
detect the leak. I am not the author of the code so it is quite hard
for me to debug it.

So I have currently for example a little main which does not leak at
all on a desktop linux version (Fedora 17) which use the same glib
version( perhaps not exactly the same verson but major, minor and
macro number version seems to be the same. But when I run this piece
of code on my embedded linux, I have a leak of 4 Bytes per loop
iteration.

Code is :
#define SIZE_INNER_ARRAY 5
...

static void
data_constructor_array_struct_string_array_struct_string_variant(GPtrArray
*data, const gchar *client_objname)
{
  guint member_count;
  GValueArray *outer_struct = g_value_array_new(1);
  GPtrArray *inner_array = g_ptr_array_new ();

  printf ("client objname : %s\n", client_objname);

  g_value_array_append(outer_struct, NULL);
  g_value_init(g_value_array_get_nth(outer_struct, 0), G_TYPE_STRING);
  g_value_set_string(g_value_array_get_nth(outer_struct, 0),
client_objname);

  for(member_count=0;member_count < SIZE_INNER_ARRAY ;member_count++)
  {
    GValueArray *inner_struct = g_value_array_new(2);
    g_value_array_append(inner_struct, NULL);
    g_value_init(g_value_array_get_nth(inner_struct, 0), G_TYPE_STRING);

    char str[32];
    sprintf (str, "TUTU%d", member_count);
    g_value_set_string(g_value_array_get_nth(inner_struct, 0), str);

    GValue *ptr = g_new0 (GValue, 1);
    g_value_init (ptr, G_TYPE_UINT);
    g_value_set_uint (ptr, member_count);

    g_value_array_append(inner_struct, NULL);
    g_value_init(g_value_array_get_nth(inner_struct, 1), G_TYPE_VALUE);
    g_value_set_static_boxed(g_value_array_get_nth(inner_struct, 1), ptr);
    //g_free(ptr); // can remove it, but use set_static instead of set

    g_ptr_array_add(inner_array, inner_struct);
  }

  g_value_array_append(outer_struct, NULL);
  g_value_init(g_value_array_get_nth(outer_struct, 1),
dbus_g_type_get_collection("GPtrArray",dbus_g_type_get_struct
("GValueArray", G_TYPE_STRING, G_TYPE_VALUE, G_TYPE_INVALID)));
  g_value_set_boxed (g_value_array_get_nth(outer_struct, 1), inner_array);

  g_ptr_array_add(data, outer_struct);
}

int main (void)
{
  GPtrArray *ptr_array = NULL;
  g_type_init ();
  //while (1)
  {
    ptr_array = g_ptr_array_new ();
    data_constructor_array_struct_string_array_struct_string_variant
(ptr_array, "mon objet");
    //sleep (3);
  }
  return 0;
}

To summarize the code. I have an outer_struct which is a GValueArray.
First element is a gvalue with a string. Second element is inner_array
which is a GPtrArray. Each element of inner_array is a innner_struct
which is a GValueArray compose of two elements. As you can see in this
example I am adding 5 inner_struct in the inner_array.

Do I am doing something wrong in this code ? Do I have a memory leak
in my glib version ?
I hope that it is the good place to ask my request.

Thanks & Regards
Arthur.
_______________________________________________
gnome-devel-list mailing list
gnome-devel-list gnome org
https://mail.gnome.org/mailman/listinfo/gnome-devel-list





--
- Arthur LAMBERT



-- 
- Arthur LAMBERT


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]