[Webkit-unassigned] [Bug 32197] feDiffuseLighting filter is not implemented

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=32197





--- Comment #10 from Dirk Schulze <krit at webkit.org>  2010-05-03 11:40:01 PST ---
(In reply to comment #9)
> This is just a draft patch, not a finished one, so ChangeLog and build systems
> can be wait. What I need at this stage is the guidance of a reviewer :) )As
> Darin said before)
> 
> The question here is do you like my appoach, or you would prefer something
> else. In my patch, I combind the diffuse and specular lighting interfaces to a
> common base class. This base class contains all common, and own arguments. It
> makes the implementation much easier, but has a space overhead. I think this
> overhead is acceptable because of the low number of such objects. Do you agree?
> I mean you said "I should think about...", but you are the reviewer not me :)
> Would you prefer a patch doing that? Is that fit for the current implementation
> style or not?
> 
> The other thing is how the lighting source interface is changed. I have added
> members to the interface which is used when the painting happens. This is not a
> thread safe approach, but faster than passing the same arguments again and
> again. This is a balance question, do you prefer speed or thread safe
> operation?
> 
> I put this patch here, because only the diffuse lighting part is finished,
> except one thing: I don't understand the purpose of kernel units. The official
> definition is fuzzy, and firefox results for different kernel units are
> strange. It makes the image "dirty", I can't describe it better. Lots of random
> pixels appear. I thought it is kind of a scaling argument, but I am not sure
> anymore.
> 
> Also interesting, that the cut-off argument of spot light is not used by the
> spot light apply formulas. I could create rules for that, but not sure that is
> the best way, since I should follow the rules in the standard.
> 
> 
> > Don't we have constants for M_PI / 180?
> Where? Or where to put DegreeToRad?
> 
> First I plan to finish the code, the pixel tests can wait.

The problem on combining the both effects is on changing the effects (by script
or animation). We'll recreate the complete filter on changing one effect, but
this should change in the future and only effects that need updates should be
rerendered. So we might leave the FEDiffuseLighting.cpp.
For the graphic side of implementation, you should ask a graphic guy. Oliver
Hunt and others have experience here.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.