WhatsCanvas GL-Family Blend Mode Audit

This audit records how public Paint::BlendMode values map to the current GL-family backend.

Current Pipeline Assumption

• Shader outputs are treated as non-premultiplied RGBA.
• RenderContext::applyBlendMode enables GL_BLEND and uses glBlendEquation(GL_FUNC_ADD).
• Blend state is applied per command before draw execution.
• Points, lines, paths, images, text, and saveLayer restore commands all snapshot a DrawBlendMode.
• SpriteBatch batches compatible image commands with the command blend mode preserved.
• There is no public premultiplied-alpha output option today.

Public To Renderer Mapping

Public paint mode	Renderer mode	GL RGB factors	GL alpha factors	Status
SRC_OVER	SrcOver	SRC_ALPHA, ONE_MINUS_SRC_ALPHA	ONE, ONE_MINUS_SRC_ALPHA	Standard default
SRC	Src	ONE, ZERO	ONE, ZERO	Replaces destination
DST	Dst	ZERO, ONE	ZERO, ONE	Keeps destination
CLEAR	Clear	ZERO, ZERO	ZERO, ZERO	Clears affected pixels
SRC_IN	SrcIn	DST_ALPHA, ZERO	DST_ALPHA, ZERO	Porter-Duff approximation
DST_IN	DstIn	ZERO, SRC_ALPHA	ZERO, SRC_ALPHA	Porter-Duff approximation
SRC_OUT	SrcOut	ONE_MINUS_DST_ALPHA, ZERO	ONE_MINUS_DST_ALPHA, ZERO	Porter-Duff approximation
DST_OUT	DstOut	ZERO, ONE_MINUS_SRC_ALPHA	ZERO, ONE_MINUS_SRC_ALPHA	Porter-Duff approximation
SRC_ATOP	SrcAtop	DST_ALPHA, ONE_MINUS_SRC_ALPHA	DST_ALPHA, ONE_MINUS_SRC_ALPHA	Porter-Duff approximation
DST_ATOP	DstAtop	ONE_MINUS_DST_ALPHA, SRC_ALPHA	ONE_MINUS_DST_ALPHA, SRC_ALPHA	Porter-Duff approximation
XOR	Xor	ONE_MINUS_DST_ALPHA, ONE_MINUS_SRC_ALPHA	ONE_MINUS_DST_ALPHA, ONE_MINUS_SRC_ALPHA	Porter-Duff approximation
ADD	Add	SRC_ALPHA, ONE	ONE, ONE	Additive
MULTIPLY	Multiply	DST_COLOR, ZERO	DST_ALPHA, ZERO	Fixed-function approximation
SCREEN	Screen	ONE, ONE_MINUS_SRC_COLOR	ONE, ONE_MINUS_SRC_ALPHA	Fixed-function approximation

Important Limitations

• These modes rely on fixed-function blending. They are not a full shader-composited blend stack.
• Multiply and screen are approximate for the current non-premultiplied outputs.
• Advanced Photoshop-style modes are not implemented.
• A premultiplied-alpha option would need a separate public contract and shader/output audit.
• Exact visual results can vary with framebuffer format and driver behavior, so blend-heavy scenes are good candidates for fuzzy visual comparison.

Validation Coverage

• PaintStateTests verifies public blend-mode state round trips.
• The showcase demo exercises ADD, MULTIPLY, SCREEN, SRC_IN, and DST_OUT.
• The validation scene suite includes blend usage in gradient/effect and saveLayer scenes.
• scripts/compare_ppm_fuzzy.py can be used for blend-heavy visual baselines where exact hashes are too strict.