| Message |
 |
- Date: 2012-06-05 02:59
- Sender: Daniel Stokes
- I am not sure how you created your patch, but I had to tweak the file a little to get it to apply right.
More importantly though, if you edit the python API make sure to edit the docs corresponding to that part of the API. This is doable in the appropriate rst file in the doc/python_api folder.
|
- Date: 2012-07-29 02:12
- Sender: Mitchell Stokes
- There is a memory leak: KX_GameObject::m_pInstanceObjects is never cleaned up.
|
- Date: 2012-07-29 02:37
- Sender: Mitchell Stokes
- Okay, I've uploaded a version of the patch with better refcounting. Visual Leak Detector is not reporting any more leaks with this version. :)
|
- Date: 2012-07-29 02:42
- Sender: Mitchell Stokes
- I just noticed that you have added some KX_PYMETHODs in KX_GameObject.h that are never actually getting used, and can be removed.
|
- Date: 2012-07-30 01:45
- Sender: Mitchell Stokes
- Looks like the improved patch can still cause leaks with replica objects. Some additional processing might have to be done in KX_GameObject::GetReplica().
|
- Date: 2012-08-06 20:28
- Sender: Martin Sell
- Your patch causes a zombie object for every group proxy, that is deleted.
I have added a check for group objects in newDeleteObject():
If the group object is deleted, all members should loose their reference to the group object.
+ if(newobj->GetInstanceObjects()) {
+ for (int i = 0; i < newobj->GetInstanceObjects()->GetCount(); i++) {
+ KX_GameObject* instance = (KX_GameObject*)newobj->GetInstanceObjects()->GetValue(i);
+ instance->RemoveDupliGroupObject();
+ }
+ }
I still have some stuff to cleanup, an assertion is not needed in RemoveDupliGroupObject() and I will also cleanup the KX_PYMETHODs.
|
- Date: 2012-08-06 20:34
- Sender: Martin Sell
- Tried to upload, the patch, did not work the first time..
|
- Date: 2012-09-17 09:48
- Sender: Mitchell Stokes
- I've been using this patch quite a bit in my current game project, and I'm not getting any more zombie objects. I'm attaching the copy of the patch that I've been using. It's mostly the same as Martin's last patch, but I've removed an assert (it was right in front of a NULL check anyways), and I made the patch with TortoiseSVN.
I'm assigning this to Campbell for final review/okay.
|
- Date: 2012-09-19 20:32
- Sender: Martin Sell
- Thanks for your support, Mitchell!
|
- Date: 2012-10-07 02:10
- Sender: Mitchell Stokes
- This patch has now been committed as r51131.
Congrats on finally getting this into trunk!
|
