[Soc-2010-dev] Status Report
leif.a.andersen at gmail.com
Fri May 28 19:59:03 CEST 2010
This week went okay for the first week. I'm not really behind, although I
would like to go faster next week.
1. Tested out several frameworks (by building sample tests)
iii. Hashcode tests (using the pyunit framework to aggregate the tests).
2. Started developing a python test module, which can be found by using
i. Module contains 3 parts
ii. These are essentially helper methods for getting a hashcode(), or
clearing all of the scenes/objects/meshes
3. Aggregated all of the tests using CTest.
i. Currently though, all of the pyunit tests have to pass regular
expression OK to pass. That could potentially be problematic, but to return
codes to work, I would have to break the python script, by making failures
return different codes.
ii. GTest based tests work fine with return codes.
iii. Currently, all tests are added in their directories CMakeLists.txt
file, perhaps it would be prudent to add them all to one file, and leave
those directories out of the project.
iv. Also, the python based test rely upon typing in a command, which
assumes the user hasn't moved the source directory.
v. make test can be run to run the test suite demo, alternatively, running
ctest in the test directory will run the tests without making anything.
4. Did preliminary work with PIL, which also looks promising for comparing
scenes as apposed to hash codes.
5. Took a look at using CDash for visual aggregation to blender. Little on
this front though.
6. Caused a few debates to be made on #blendercoders about the type of
tests needed, in particular, dumb 'catch all' tests vs. a
few strategically placed tests. Still not sure which is better though, so
I'm working on both.
1. Moderate Priority - While most of the libraries of the current blender
source directory will work fine, for some reason, functions
in BLI_path_util.h, specifically BLI_char_switch() do not work properly, and
will cause the code to fail to compile. Every other tested function works.
Only fails with GCC, MSVC works fine. No other compilers tested.
Plans for next week:
1. More dabbles with GTest, some time to be spent to unresolved problem #1.
2. Adding several more helper methods to the tests module. To be useful
for automated unit test creation.
3. More dabbling into CDash
4. Removing the hard coded hashes for the hash test, make it a text file,
then possibly adding functions to change the hashes, if blender changes.
i. Research should also be done to ensure all changes were intended, and no
unintended changes crept under the radar.
5. Some clean up of the tests directory, and code in there.
6. Actually adding more tests, this should be easier after #2 is completed.
That was easy:
On Fri, May 28, 2010 at 09:24, Rohith B.V. <rohith291991 at gmail.com> wrote:
> This Week:
> 1) Edited relevant files to make a dummy modifier called "Quadrangulate".
> 2) Checked out the CoMiSo module which is used to solve mixed integer
> 3) Decided to make my own mixed integer solver, basing it off CoMiSo.
> 4) Planned on using Eigen2 for this as it already exists in the blender
> 5) Studied the code and documentation of Eigen2 and acquired thorough
> understanding of its working.
> 6) Implemented a simple sparse cholesky solver using Eigen2. Also made some
> empty functions to be filled later.
> Next Week:
> 1) Hopefully complete the Mixed Integer solver module. This includes Gauss
> Seidel, Conjugate Gradient and Cholesky modules.
> 2) Integrate above module into the above dummy "Quadrangulate" modifier.
> So far no issues. Speed has been OK but will pick up next week onwards,
> since I don't have to read too much documentation.
> Soc-2010-dev mailing list
> Soc-2010-dev at blender.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Soc-2010-dev