Windows compiler complaint cleanup

Windows compiler complaint cleanup.

Cary - are the changes in SkSetPoly3To3*.cpp what was intended?

[bsalomon, 2012-04-10 13:39:09]

Adding Stephen for the blur scalar->int texture dimensions changes.

[caryclark1, 2012-04-10 13:58:30]

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp
File experimental/SkSetPoly3To3_A.cpp (left):

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp...
experimental/SkSetPoly3To3_A.cpp:35: return (SkDScalar)a * b;
This is taking two 16.16 fixed point values and returning a 32.32 fixed point
value. The original intent of this cast was to avoid the compiler performing a
32 bit x 32 bit = intermediate 32 bit result -> promoted to 64 bit. It would
probably be safer to rewrite this as:
       return (SkDScalar) ((SkDScalar) a * b);
if the compiler allows that.

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp
File experimental/SkSetPoly3To3_A.cpp (right):

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp...
experimental/SkSetPoly3To3_A.cpp:62: bool SkSetPoly3To3_A(SkMatrix* matrix,
const SkPoint src[3], const SkPoint dst[3]) {
The original intention is that the caller would use _A or _D but never both, so
the functions could share the name. But if this fixes a compiler complaint, it's
fine -- I don't think this code is in use, even experimentally.

[Stephen White, 2012-04-10 14:08:04]

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp
File experimental/SkSetPoly3To3_A.cpp (right):

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp...
experimental/SkSetPoly3To3_A.cpp:35: return (SkDScalar)(a * b);
In SK_SCALAR_IS_FIXED, this looks like it will do a 32-bit int multiply, then
cast the result to a 64-bit int.  The old code looks like it was doing the
multiply in 64 bits.  (similarly in SK_SCALAR_IS_FLOAT:  float mult vs double
mult).  Are you sure this is what you want?

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_D.cpp
File experimental/SkSetPoly3To3_D.cpp (right):

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_D.cpp...
experimental/SkSetPoly3To3_D.cpp:20: return (SkDScalar)(a * b);
Same here.

http://codereview.appspot.com/5988072/diff/1/src/gpu/GrContext.cpp
File src/gpu/GrContext.cpp (right):

http://codereview.appspot.com/5988072/diff/1/src/gpu/GrContext.cpp#newcode2112
src/gpu/GrContext.cpp:2112: static_cast<float>(scaleFactorY));
This looks fine, but out of curiosity, what warnings do we have on that complain
about an implicit int->float conversion?

[robertphillips, 2012-04-10 14:53:51]

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp
File experimental/SkSetPoly3To3_A.cpp (left):

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp...
experimental/SkSetPoly3To3_A.cpp:35: return (SkDScalar)a * b;
On 2012/04/10 13:58:31, caryclark1 wrote:
> This is taking two 16.16 fixed point values and returning a 32.32 fixed point
> value. The original intent of this cast was to avoid the compiler performing a
> 32 bit x 32 bit = intermediate 32 bit result -> promoted to 64 bit. It would
> probably be safer to rewrite this as:
>        return (SkDScalar) ((SkDScalar) a * b);
> if the compiler allows that.

Done.

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp
File experimental/SkSetPoly3To3_A.cpp (right):

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_A.cpp...
experimental/SkSetPoly3To3_A.cpp:35: return (SkDScalar)(a * b);
On 2012/04/10 14:08:04, Stephen White wrote:
> In SK_SCALAR_IS_FIXED, this looks like it will do a 32-bit int multiply, then
> cast the result to a 64-bit int.  The old code looks like it was doing the
> multiply in 64 bits.  (similarly in SK_SCALAR_IS_FLOAT:  float mult vs double
> mult).  Are you sure this is what you want?

Done.

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_D.cpp
File experimental/SkSetPoly3To3_D.cpp (right):

http://codereview.appspot.com/5988072/diff/1/experimental/SkSetPoly3To3_D.cpp...
experimental/SkSetPoly3To3_D.cpp:20: return (SkDScalar)(a * b);
On 2012/04/10 14:08:04, Stephen White wrote:
> Same here.

Done.

http://codereview.appspot.com/5988072/diff/1/src/gpu/GrContext.cpp
File src/gpu/GrContext.cpp (right):

http://codereview.appspot.com/5988072/diff/1/src/gpu/GrContext.cpp#newcode2112
src/gpu/GrContext.cpp:2112: static_cast<float>(scaleFactorY));
warning C4244: 'argument' : conversion from 'int' to 'float', possible loss of
data

The compiler is just warning that all the digits of a large integer may not be
preserved by the conversion.

[robertphillips, 2012-04-10 18:33:10]

committed as r3644