On Thu, Jun 5, 2014 at 9:58 PM, Dirk Schulze <dschu...@adobe.com> wrote:
>
> On Jun 6, 2014, at 6:52 AM, Rik Cabanier <caban...@gmail.com> wrote:
>
> >
> >
> >
> > On Thu, Jun 5, 2014 at 9:40 PM, Dirk Schulze <dschu...@adobe.com> wrote:
> >
> > On Jun 6, 2014, at 6:27 AM, Robert O'Callahan <rob...@ocallahan.org>
> wrote:
> >
> > > On Fri, Jun 6, 2014 at 4:22 PM, Dirk Schulze <dschu...@adobe.com>
> wrote:
> > > What about
> > >
> > > DOMMatrix(1,0,0,1,0,0) or
> > > DOMMatrix(1,0,0,0,0,1,0,0,0,0,1,0,0,0,0,1)
> > >
> > > Do we check the values and determine if the matrix is identity or not?
> If we do, then authors could write DOMMatrix(other.a, other.b, other.c,
> other.d, other.e, other.f) and check any kind of matrix after transforming
> for identity. In this case, a real isIdentity check wouldn’t be worst IMO.
> > >
> > > I would lean towards just setting isIdentity to false for that case,
> but I could go either way. If authors try really hard to shoot themselves
> in the foot, they can.
> > >
> > > Rob
> >
> > Just as comparison: Gecko checks for IsIdentity 75 times (exclusive the
> definitions in matrix and matrix4x4). Every time the values are simply
> checked for 0 and 1. Means Gecko is shooting itself in the foot quite often
> :P. (In WebKit it is about ~70 times as well.)
> >
> > The question is not that 'isIdentity' is bad. Benoit's issue was that
> checking for 'isIdentity' after doing transformations might cause jittery
> results (ie switch to true or false depending on the conditions).
> > Quickle scanning mozilla's codebase, our current definition of
> 'isIdentity' would return the correct result in all cases.
> >
>
> Just take the first result of many:
>
> static PathInterpolationResult
> CanInterpolate(const SVGPathDataAndInfo& aStart,
> const SVGPathDataAndInfo& aEnd)
> {
> if (aStart.IsIdentity()) {
> return eCanInterpolate;
> }
> …
>
> Where can you guarantee that you don’t see jittering? aStart could be
> modified
>
That one does not check for an identity matrix:
/**
* Returns true if this object is an "identity" value, from the
perspective
* of SMIL. In other words, returns true until the initial value set up in
* SVGPathSegListSMILType::Init() has been changed with a SetElement()
call.
*/
bool IsIdentity() const {
if (!mElement) {
NS_ABORT_IF_FALSE(IsEmpty(), "target element propagation failure");
return true;
}
return false;
}
Maybe you and I should take this offline...
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform
