[Gnash-commit] gnash ChangeLog server/sprite_instance.cpp

[Sandro Santilli]

CVSROOT:        /sources/gnash
Module name:    gnash
Changes by:     Sandro Santilli <strk>  08/01/04 13:10:45

Modified files:
        .              : ChangeLog 
        server         : sprite_instance.cpp 

Log message:
        beginGradientFill: Improvements in non-boxed matrixType use,
        comments about future improvements.

CVSWeb URLs:
http://cvs.savannah.gnu.org/viewcvs/gnash/ChangeLog?cvsroot=gnash&r1=1.5294&r2=1.5295
http://cvs.savannah.gnu.org/viewcvs/gnash/server/sprite_instance.cpp?cvsroot=gnash&r1=1.426&r2=1.427

Patches:
Index: ChangeLog
===================================================================
RCS file: /sources/gnash/gnash/ChangeLog,v
retrieving revision 1.5294
retrieving revision 1.5295
diff -u -b -r1.5294 -r1.5295
--- ChangeLog   4 Jan 2008 12:24:31 -0000       1.5294
+++ ChangeLog   4 Jan 2008 13:10:44 -0000       1.5295
@@ -1,5 +1,11 @@
 2008-01-04 Sandro Santilli <address@hidden>
 
+       * server/sprite_instance.cpp (beginGradientFill): 
+         Improvements in non-boxed matrixType use, comments
+         about future improvements.
+
+2008-01-04 Sandro Santilli <address@hidden>
+
        * testsuite/server/MatrixTest.cpp: add a test
          for transforming a couple of points using
          a translate + scale approach. This is the

Index: server/sprite_instance.cpp
===================================================================
RCS file: /sources/gnash/gnash/server/sprite_instance.cpp,v
retrieving revision 1.426
retrieving revision 1.427
diff -u -b -r1.426 -r1.427
--- server/sprite_instance.cpp  3 Jan 2008 19:58:54 -0000       1.426
+++ server/sprite_instance.cpp  4 Jan 2008 13:10:45 -0000       1.427
@@ -1346,7 +1346,21 @@
        // Parse matrix
        // ----------------------------
        
-       // TODO: fix this matrix build-up, it is NOT correct
+       //
+       // TODO: fix the matrix build-up, it is NOT correct for
+       //       rotation.
+       //       For the "boxed" matrixType and radial fills this
+       //       is not a problem as this code just discards the
+       //       rotation (which doesn't make sense), but for
+       //       the explicit matrix type (a..i) it is a problem.
+       //       The whole code can likely be simplified by 
+       //       always transforming the gnash gradients to the
+       //       expected gradients and subsequently applying
+       //       user-specified matrix; for 'boxed' matrixType
+       //       this simplification would increas cost, but
+       //       it's too early to apply optimizations to the
+       //       code (correctness first!!).
+       //
 
        matrix mat;
        matrix input_matrix;
@@ -1394,6 +1408,7 @@
                        // For linear gradients, dunno why translation must be 
negative...
                        input_matrix.concatenate_translation( -valX, -valY );
                        //cout << "inpt matrix with concatenated translation: " 
<< input_matrix << endl;
+
                }
 
                mat.concatenate(input_matrix);
@@ -1415,15 +1430,52 @@
                float valG = 
PIXELS_TO_TWIPS(matrixArg->getMember(keyG).to_number()); // x0
                float valH = 
PIXELS_TO_TWIPS(matrixArg->getMember(keyH).to_number()); // y0
 
-               input_matrix.m_[0][0] = valA;
-               input_matrix.m_[1][0] = valB;
-               input_matrix.m_[0][1] = valD;
-               input_matrix.m_[1][1] = valE;
-               input_matrix.m_[0][2] = valG;
-               input_matrix.m_[1][2] = valH;
+               input_matrix.m_[0][0] = valA; // xx
+               input_matrix.m_[1][0] = valB; // yx
+               input_matrix.m_[0][1] = valD; // xy
+               input_matrix.m_[1][1] = valE; // yy
+               input_matrix.m_[0][2] = valG; // x0
+               input_matrix.m_[1][2] = valH; // y0
+
+               // This is the matrix that would transform the gnash
+               // gradient to the expected flash gradient.
+               // Transformation is different for linear and radial
+               // gradient for Gnash (in flash they should be the same)
+               matrix gnashToFlash;
+
+               if ( radial )
+               {
+
+                       // Gnash radial gradients are 64x64 with center at 32,32
+                       // Should be 20x20 with center at 0,0
+                       float g2fs = 20.0/64.0; // gnash to flash scale
+                       gnashToFlash.set_scale(g2fs, g2fs);
+                       gnashToFlash.concatenate_translation(-32.0, -32.0);
+
+               }
+               else
+               {
+                       // First define a matrix that would transform
+                       // the gnash gradient to the expected flash gradient:
+                       // this means translating our gradient to put the
+                       // center of gradient at 0,0 and then scale it to
+                       // have a size of 20x20 instead of 256x1 as it is
+                       //
+                       // Gnash linear gradients are 256x1 with center at 128,0
+                       // Should be 20x20 with center at 0,0
+                       gnashToFlash.set_scale(20.0/256.0, 20.0/1);
+                       gnashToFlash.concatenate_translation(-128.0, 0.0);
+
+               }
+
+               // Apply gnash to flash matrix before user-defined one
+               input_matrix.concatenate(gnashToFlash);
+
+               // Finally, and don't know why, take
+               // the inverse of the resulting matrix as
+               // the one which would be used
+               mat.set_inverse( input_matrix );
 
-               mat.concatenate_scale(20/16384.0);
-               mat.concatenate(input_matrix);
        }
 
        //cout << mat << endl;