[Soc-2008-dev] Friday Reports!

[Maxime Curioni]

Hello everyone,

Here is my week 1 report. If you want more details on my development
progress, I try to post updates regularly on my blog:
http://maximecurioni.com/gsoc2008. If you have remarks or suggestions,
please let me know. My latest revision is 15047.


a) What you did in the previous week;

This week, I completed the most important part of the first phase, creating
the first Freestyle-based renders within Blender. Up to last week, I had
only managed to open the 3ds file with lib3ds and hadn't gotten the view map
calculations working. The basic execution process is divided in these few
steps:

   1. instantiate the config path, the controller and the view
   2. set the controller's view to the instantiated view and set up the view
dimensions
   3. load a 3ds file
   4. insert a style module
   5. compute the view map
   6. draw the strokes
   7. render the image

 The original Freestyle program uses a Qt widget QGLViewer to display and
manipulate 3D objects (step 6 and 7). I found out this week that the
widget's functionality is also needed for the view map calculations (step
5). I managed to modify QGLViewer's code (which is open source) to only keep
the strict necessary for the job. After a few trials, I got it working.

The strokes are rendered using commands in a style module, coded in Python.
Python commands are executed through a class called PythonInterpreter, that
routes calls to Python's C interface (PyRun_SimpleString and PyErr).
Unfortunately, these methods simply crash, probably due to Blender's Python
environment. Instead of fighting against the system in place, I decided to
do some phase 2 work, having Freestyle use Blender's Python environment
directly for its execution. It just worked. That is going to save me time in
the next phase.

Once the Python commands were sent, I had to make the connection between
Python and Freestyle possible, which is a SWIG wrapper. I managed to get
that working: I compiled the SWIG wrapper as a shared library, used by the
Python runtime. I haven't found a way to generate that library automatically
with scons so right now, the wrapper module is compiled and linked manually
(instructions are in the source/blender/freestyle/SConscript file)

Once that was working, I corrected a few integration bugs related to
replacing Qt image library with ImBuf and got to a first render. I thought I
would have to play around with the OpenGL context but that seems to work by
itself (but it only works if "Render Display" is set to "Render Window"). My
first render was incomplete (colors were inverted) but I realized that I had
only gone up to step 6. By using the view's drawing method, the render
became now correct (see http://maximecurioni.com/gsoc2008/?p=18 for
examples).


b) What you plan to do in the following week;

Right now, the result is limited to a static scene and a static style module
(compiled in code). The next step is removing this fixed information and
rendering the scene model, using the camera information. I will use
Blender's 3ds export capabilities to produce a 3ds file reimported later
into Freestyle. I will also need to set up a Freestyle render panel to
control the different options used by the view.

Before I do that, I want to make sure my work compiles and runs well on
different platforms. Jean-Luc has tried compiling it and is facing issues
with SWIG. I also had mfoxdogg help me compile it on Linux but didn't have a
chance to finish it. I hope to do that in the next few days.


c) Describe any issues you are having that may require extra attention;

On my setup, I have no particular issue but I need to make sure that's also
the case for other people. I also have had a crash with a particular 3ds
file during view map calculations. I am not sure whether it is a problem
with Freestyle's algorithm, lib3ds or something else. Once I get the scene
model imported, I'll test it again. Stéphane (one of the original Freestyle
authors) also mentioned that the image aspect ratio is maybe wrong. I set
the view's dimensions to the render window's; I will have to determine
whether OpenGL projection matrices need to be tweaked.


d) Let us know whether you feel that you are on schedule, ahead of schedule,
or behind schedule\

So far, I feel I am a little bit ahead of schedule. I had planned to have my
first render by the end of this week-end (the first phase is due for June
8). Even though that's a good thing, I probably spent more time than I
should have and I am running a bit late in some of my school work. I will
try to get some of it covered this week-end.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/soc-2008-dev/attachments/20080530/5735d2f5/attachment.htm