Add GrRenderTarget parameter to GrTextContext::canDraw

src/gpu/GrTextContext.cpp

Index: src/gpu/GrTextContext.cpp
diff --git a/src/gpu/GrTextContext.cpp b/src/gpu/GrTextContext.cpp
index ad5e7c0aa5e1e21347641625d5fb982895adbbe9..13ae49166ec32f1484c3b2f2fc7cee153ef8f1af 100644
--- a/src/gpu/GrTextContext.cpp
+++ b/src/gpu/GrTextContext.cpp
@@ -46,7 +46,7 @@ bool GrTextContext::drawText(GrRenderTarget* rt, const GrClip& clip, const GrPai
 
     GrTextContext* textContext = this;
     do {

[vbuzinov, 2015/03/17 13:06:56]

Can we create a pipeline builder here and pass it to canDraw instead of rt (rt
can be queried with getRenderTarget)?  That would allow us to just use the same
code in canDraw as in GrStencilAndCoverPathRenderer.  

We could then pass it to onDrawText as well to the implementation that can draw.
 Unless its creation has to be deferred until later for some reason?

[bsalomon, 2015/03/17 14:45:48]

I like that direction (passing a PB here). I know there were some complexities
that caused us to pass both a SkPaint and GrPaint here. IIRC it is related to
LCD text. The XP work was supposed to help simplify that. It'd be great to just
turn the GrPaint into a GrPB here and drop the two paints. But I'm not familiar
with the details. Adding Greg and Jim to comment.

-        if (textContext->canDraw(skPaint, viewMatrix)) {
+        if (textContext->canDraw(rt, skPaint, viewMatrix)) {
             textContext->onDrawText(rt, clip, paint, skPaint, viewMatrix, text, byteLength, x, y);
             return true;
         }
@@ -67,7 +67,7 @@ bool GrTextContext::drawPosText(GrRenderTarget* rt, const GrClip& clip, const Gr
 
     GrTextContext* textContext = this;
     do {
-        if (textContext->canDraw(skPaint, viewMatrix)) {
+        if (textContext->canDraw(rt, skPaint, viewMatrix)) {
             textContext->onDrawPosText(rt, clip, paint, skPaint, viewMatrix, text, byteLength, pos,
                                        scalarsPerPosition, offset);
             return true;