memory leak in python process

Can someone confirm this? Pascal Sachs, have you tried the new release (0.36)? If not, please have a try and tell us if the bug is still happening.

Hi

I tried it with 0.36. I now have another configuration because not all
of the widgets are still available, but there is some improvement. The
constant memory leak is gone, so the gDesklets is usable now. But what I
recognized, is that after each reload of a widget the memory usage is
growing. Because this is not really a thing one often do, it is not that
bad, but I think this should be avoidable.

Cheers Pascal Sachs

Am Dienstag, den 04.03.2008, 01:35 +0000 schrieb MarioGonzalez:
> Can someone confirm this? Pascal Sachs, have you tried the new release
> (0.36)? If not, please have a try and tell us if the bug is still
> happening.
>

This bug has been fixed in the 0.36 release.

El mar, 04-03-2008 a las 15:36 +0000, Pascal Sachs escribió:
> But what I
> recognized, is that after each reload of a widget the memory usage is
> growing. Because this is not really a thing one often do, it is not

 Yes, I can confirm this since 0.35v or after. Can you report this as a
"New Bug" please?

> that
> bad, but I think this should be avoidable.
--
www.mgonzalez.cl

That's sometimes due to pygtk or any other libs. Nothing has been changed so
far wrt. the C code in gdesklets.

What's the status on this one ?
Can it be closed ?

This bug should stay open because of
https://bugzilla.redhat.com/show_bug.cgi?id=747420

In my opinion, this bug was introduced with revision 142!
A fix in utils/render.c creates copies of the GdkPixbuf which in turn get free'd only, if the height or width of the area drawn changes. I created a patch for this issue that solved the problem for me (running rev. 186)

Here is the patch:

--- ../gdesklets/utils/render.c 2010-05-24 12:56:05.000000000 +0200
+++ utils/render.c 2012-01-30 10:56:46.163768331 +0100
@@ -89,7 +89,10 @@
   /* set image */
   gtk_image_set_from_pixbuf (image, scaled);

+// this condition is causing a memory leak since rev. 142
+/*
   if (srcwidth != width || srcheight != height)
+*/
     g_object_unref (scaled);
 }

Can anyone confirm this?

Where is the patch?