Add gm that tests shaded stroked rectangles.

gm/stroke_rect_shader.cpp

Index: gm/stroke_rect_shader.cpp
diff --git a/gm/stroke_rect_shader.cpp b/gm/stroke_rect_shader.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..99cf38766369642b1a1a68773b97878bed08a5e4
--- /dev/null
+++ b/gm/stroke_rect_shader.cpp
@@ -0,0 +1,60 @@
+/*

[robertphillips, 2016/07/06 15:01:47]

What's up with the naming of this file?

[bsalomon, 2016/07/06 15:57:53]

I don't understand. What is up with it? The underscores?

$ ls  *_*
all_bitmap_configs.cpp      degenerate_rrect.cpp            gm_json.pyc         
      image_pict.cpp                    repeated_bitmap.cpp
annotated_text.cpp          draw_bitmap_rect_skbug4374.cpp  gradient_matrix.cpp 
      image_shader.cpp                  skbug_257.cpp
clip_strokerect.cpp         glyph_pos_align.cpp            
gradients_2pt_conical.cpp  internal_links.cpp                skbug_4868.cpp
complexclip_blur_tiled.cpp  glyph_pos.cpp                  
gradients_no_texture.cpp   path_stroke_with_zero_length.cpp 
stroke_rect_shader.cpp
convex_all_line_paths.cpp   gm_expectations.h              
hardstop_gradients.cpp     pdf_never_embed.cpp              
tilemodes_scaled.cpp

+ * Copyright 2016 Google Inc.
+ *
+ * Use of this source code is governed by a BSD-style license that can be
+ * found in the LICENSE file.
+ */
+
+#include "gm.h"
+#include "SkGradientShader.h"
+

[robertphillips, 2016/07/06 15:01:47]

Are we testing the degenerate rect cases somewhere else ?

[bsalomon, 2016/07/06 15:57:53]

The strokedrect GM tests more varieties of rects + stroking params including
degenerate rectangles.

+namespace skiagm {
+

[robertphillips, 2016/07/06 15:01:47]

// Draw stroked rects (both AA and nonAA) with all the types of joins:
//    bevel, miter, miter-limited-to-bevel, round
// and as a hairline.

[bsalomon, 2016/07/06 15:57:53]

Done.

+DEF_SIMPLE_GM(stroke_rect_shader, canvas, 690, 300) {
+    static constexpr SkRect kRect {0, 0, 100, 100};
+    static constexpr SkPoint kPts[] {{kRect.fLeft, kRect.fTop}, {kRect.fRight, kRect.fBottom}};
+    static constexpr SkColor kColors[] {SK_ColorRED, SK_ColorBLUE};
+    SkPaint paint;

[robertphillips, 2016/07/06 15:01:47]

What are your thoughts on the style proposal that we never hide sk_sp in an auto
(since it obscures ownership transactions?

[bsalomon, 2016/07/06 15:57:53]

I don't agree with it, but I'm fine abiding by it.

+    auto shader = SkGradientShader::MakeLinear(kPts, kColors, nullptr, 2,
+                                               SkShader::kClamp_TileMode);
+    paint.setShader(std::move(shader));
+    paint.setStyle(SkPaint::kStroke_Style);
+    // Do a large initial translate so that local coords disagree with device coords significantly
+    // for the first rect drawn.

[robertphillips, 2016/07/06 15:01:47]

kRect.centerX(), kRect.centerY() ?

[bsalomon, 2016/07/06 15:57:53]

Done.

+    canvas->translate(kRect.width() / 2, kRect.height() / 2);
+    static constexpr SkScalar kPad = 20;
+    for (auto aa : {false, true}) {
+        paint.setAntiAlias(aa);
+        canvas->save();
+
+        static constexpr SkScalar kStrokeWidth = 10;
+        paint.setStrokeWidth(kStrokeWidth);
+
+        paint.setStrokeJoin(SkPaint::kBevel_Join);
+        canvas->drawRect(kRect, paint);
+        canvas->translate(kRect.width() + kPad, 0);
+
+        paint.setStrokeJoin(SkPaint::kMiter_Join);
+        canvas->drawRect(kRect, paint);
+        canvas->translate(kRect.width() + kPad, 0);
+
+        // This miter limit should effectively produce a bevel join.
+        paint.setStrokeMiter(0.01);
+        canvas->drawRect(kRect, paint);
+        canvas->translate(kRect.width() + kPad, 0);
+

[robertphillips, 2016/07/06 15:01:47]

Wrong comment ?

[bsalomon, 2016/07/06 15:57:53]

Done.

+        // This miter limit should effectively produce a bevel join.
+        paint.setStrokeJoin(SkPaint::kRound_Join);
+        canvas->drawRect(kRect, paint);
+        canvas->translate(kRect.width() + kPad, 0);
+

[robertphillips, 2016/07/06 15:01:47]

// Hairlines, make bevel and round joins appear the same as mitered ones

Do we want to test that ?

[bsalomon, 2016/07/06 15:57:53]

If we are going to test that I think it'd be better to do so in a GM focused on
hairline rects  (if we don't have one already). Or it could be done as a unit
test that just asserts that join doesn't affect the generated pixels.

+        paint.setStrokeWidth(0);
+        canvas->drawRect(kRect, paint);

[robertphillips, 2016/07/06 15:01:47]

Rm this translate ?

[bsalomon, 2016/07/06 15:57:53]

Done.

+        canvas->translate(kRect.width() + kPad, 0);
+
+        canvas->restore();
+        canvas->translate(0, kRect.height() + kPad);
+    }
+}
+
+}