This would fail building with Visual Studio 2017, at least downstream in Godot
where we undefine old Windows compilers' non-standard `min`/`max` macros (see
`minmax.h`/`NOMINMAX`).
Some gcc versions and configurations was giving an error:
error: "comparison of integer expressions of different signedness: 'const SwCoord'
{aka 'const long int'} and 'uint32_t' {aka 'unsigned int'} [-Werror=sign-compare]"
This patch should fix the error
Introduced CompositeMethod::LumaMask that converts the source pixels to the
grayscale (luma value) before alpha blending. Thanks to it, mask works more like
typical mask in graphics editor software.
Grayscale is calculated with weighted method:
(0.0721*B + 0.7154*G + 0.2125*R) * A
Introduced surface->blender.lumaValue function
Handle correctly duplicated spans from the multiple y span data.
Previous logic only expected the one single y span data from the rle.
However rle might have multiple y span data if the anti-aliasing is applied.
This patch also removed the bad design of the common engine
which handles the anti-alising option to ignore the anti-aliasing rle generation.
Just realized, it's difficult to control that condition due to scene-composition.
strokeOutline returned strokeExportOutline() is the address of mpool->strokeOutline[idx].
Assuming this value is null, mpoolRetStrokeOutline on line 617 will access mpool->strokeOutline[idx].
Logically, the logic of the mpool* functions does not occur in this case.
replace the transformed rle rgba image with the texmap raster.
this patch removes the several scattered transformed image rasterizer,
reuse the unified one texmap method instead.
yay, it's much clean and optimal.
Previously, translucent png images are not displayed properly
due to alpha channels premultiplication.
This patch implements that missing part to support it properly
by introducing the Surface data between canvas engine & rasterizer
@Issue: https://github.com/Samsung/thorvg/issues/655
The x and y of spans cannot be negative
because they are specified as coordinates inside the buffer.
Change the type to fix warnings and potential problems
that occur in conversion between int16_t and uint32_t.
Introduced the texture-mapping algorithm for the image-scaler.
With this approach, we can support any arbitrary forms of the transformed shapes
while appling the scale factors - x/y separately.
Also this is more efficient when the transformed image is far from the rectangular form
because our current normal image rasterizer based on the given drawing region whereas
our new approach is span-based.
Thus, we applied this mechanism only for the transformed images.
There was a missing sorting between the left-top & right-bottom corner.
that results in the negative values of the viewport...
Now fixed it.
+ refactored to use math functions...
+ still it's a buggy, sometimes no draw at 90' in stress. don't know reason. :(
