Direct3D testing project (GSoC)

[Roderick Colenbrander]

On Wed, Apr 7, 2010 at 4:21 AM, Conrad Buck <conartist6 at gmail.com> wrote:
> Hello devels,
>      I'm looking at submitting a SoC proposal to the wine project, but I'm
> feeling a little overwhelmed as my current coding experience is all over the
> place, and very little of it working with existing projects. My most recent
> work has been looking  low-level code and algorithms behind basic OpenGL, so
> I thought perhaps I might be able to make relevant contributions to the
> project of testing and improving Direct3D support. Can anyone that knows
> more about whats under the hood for that sort of functionality in wine tell
> me what kind of magnitude of project I might be looking at? Only a limited
> number of bug reports have been filed, but all seem to be indicative of
> fairly major issues. Once I get more of an idea of what I might be dealing
> with (and examine some of the problems myself) I can fine-tune my proposal.
> Thanks,
>     Conrad
>
>
>

Hi,

First of all welcome to Wine. Our Direct3D -> OpenGL code is quite
complex and has a steep learning curve, so just hacking on Direct3D to
improve big chunks of code isn't that suited. That said there is
certainly work which you could do in the testing / bug hunting area.

For instance you could work on improving our test suite. Currently
mostly Nvidia systems pass the tests. For a part this is due to driver
bugs in the other drivers but some tests are also too strict and there
might be some WineD3D bugs. You could fix broken tests, fix bugs in
our Direct3D code found by the tests, you could add new tests and
perhaps you could also improve the tests on Windows. If you want you
might be able to fix some normal Direct3D bug reports as well (those
will likely require new tests). It might be hard to write a good
proposal though (especially judging whether you reached your goals
might be hard).

Another form of testing which you could improve is our 'gaming' test
suite in which performance in games is tested for regressions. I have
never used those tests but according to Stefan the test suite can use
improvements. Perhaps some visual tests could be added to them as
well, to see if there are no visual regressions. Not sure how that
should work but I guess you would need a mechanism for recording
keyboard/mouse movements and playing them back, so that you reach
fixed spots at which you can make screenshots for comparisons.

The two suggestions and some more info can also be found in a recent
thread which can be found here:
http://www.winehq.org/pipermail/wine-devel/2010-March/082698.html
http://www.winehq.org/pipermail/wine-devel/2010-March/082707.html

Good luck writing a proposal and if you have questions just ask!
Roderick