[Bf-python] BUG PYTHON identified! Please fix it...

[Willian Padovani Germano]

Hi,

----- Original Message -----
From: "jms" <jmsoler at free.fr>
To: <bf-python at blender.org>
Sent: Saturday, May 08, 2004 7:01 AM
Subject: Re: [Bf-python] BUG PYTHON identified! Please fix it...


> Willian Padovani Germano wrote:
>
> >Hi,
> >
> >Yes Manuel, you identified a memory leak, shouldn't be hard to fix, we'll do
it.
> >And the problem pointed by jms, too.
> >
>
> Note that there is a problem on mesh  with aramture and material too.

I made tests using a script like you sent, only difference was that I used the
default cube instead of a plane, and parented it in the ways available with
ctrl+p (both armature+object and object+armature, all extra options).  But I
couldn't reproduce the problem.  Yesterday, though, while testing with a Suzanne
with almost 8000 verts with the script sent by Manuel, I saw the disappearing
problem.  Could you please email me a sample .blend file that reproduces the
problem, jms?

About the crash when extra verts are added to a mesh with vertex groups, this
could be solved with a simple hack, but wouldn't be good, wouldn't work for when
verts are removed, for example.

Joseph: so it's better to fix it properly by adding the dvert info to the
nmverts, like is done for the editverts.  This way verts can be added or removed
and the info about groups won't be lost if available.  With the addition, we'll
probably rewrite some of those vgroup methods you added (and Jordi's
getVertexInfluences) to use the new simpler way.  We need to discuss a little
the internals here, like if we should also let bones be accessed from the verts
(returning a pybone).

I'll work on this, but first I need to know if a new release (2.33a maybe) will
come soon.  Will ask tomorrow at the meeting.  The changes mess with many
functions in NMesh, so we need time to finish / test them properly.  I'll commit
today the fix for the memory leak, though, it's just a single line addition.

--
Willian, wgermano at ig.com.br