710521 - Refactor gfxFont to separate out drawing stroke and drawing to path

Status: RESOLVED FIXED

Description

[Edwin Flores [inactive from 2016-12-01] [:eflores] [:edwin]]

Attachment: Patch

Comment 1

[Robert O'Callahan (:roc) (email my personal email if necessary)]

+            mTextRun->Draw(mThebes,
+                           point,
+                           gfxFont::GLYPH_STROKE,
+                           0,
+                           mTextRun->GetLength(),
+                           nsnull,
+                           nsnull);
         else
             // mOp == TEXT_DRAW_OPERATION_FILL
             mTextRun->Draw(mThebes,
                            point,
+                           gfxFont::GLYPH_FILL,
                            0,
                            mTextRun->GetLength(),
                            nsnull,
                            nsnull);

Combine these into a single Draw call with a conditional expression for the flags.

+        if (aDrawMode == gfxFont::GLYPH_FILL)
+            DrawPartialLigature(font, aContext, start, ligatureRunStart, &pt, aProvider);

+        if (aDrawMode == gfxFont::GLYPH_FILL)
+            DrawPartialLigature(font, aContext, ligatureRunEnd, end, &pt, aProvider);

{} around if body

Not being able to draw partial ligatures with stroking sucks a bit, but we'll deal I guess. We could pass the draw mode into DrawPartialLigature, which draws the glyph with some clipping to cut it up into pieces, but that wouldn't work well with stroking.

+    typedef enum {
+        GLYPH_FILL      = 1,
+        GLYPH_STROKE    = 2,
+        GLYPH_PATH      = 4
+    } DrawMode;

Need to document that GLYPH_FILL and GLYPH_STROKE can be used together, but they can't be used with GLYPH_PATH. Actually, add NS_ASSERTIONs about that to gfxTextRun:::Draw and gfxFont::Draw. Also document that GLYPH_STROKE destroys the current path in the context.

Otherwise looks good!! Please post a new patch.

Comment 2

[Robert O'Callahan (:roc) (email my personal email if necessary)]

+        GLYPH_FILL      = 1,
+        GLYPH_STROKE    = 2,
+        GLYPH_PATH      = 4

Also the indenting before = is unnecessary.

Comment 3

[Robert O'Callahan (:roc) (email my personal email if necessary)]

CCing font guys for feedback.

This is a relatively simple change to avoid using text-to-path when we just need to stroke the text. We're doing this so that when we introduce fonts with SVG glyphs, stroking can work well. (It's difficult/impossible to get a path for an SVG glyph in general.)

Comment 4

[Edwin Flores [inactive from 2016-12-01] [:eflores] [:edwin]]

Attachment: Patch #2

Comment 5

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 581791 [details] [diff] [review]
Patch #2

Review of attachment 581791 [details] [diff] [review]:
-----------------------------------------------------------------

Good stuff!

::: content/canvas/src/nsCanvasRenderingContext2D.cpp
@@ +2796,5 @@
>          }
>  
> +        mTextRun->Draw(mThebes,
> +                       point,
> +                       mOp == nsCanvasRenderingContext2D::TEXT_DRAW_OPERATION_STROKE ? gfxFont::GLYPH_STROKE : gfxFont::GLYPH_FILL,

80 column line limit

Comment 6

[Robert O'Callahan (:roc) (email my personal email if necessary)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/2674bffd5575

Comment 7

[Matt Brubeck (:mbrubeck)]

Backed out on inbound (along with other patches in the same push) because of Mac and Win7 reftest failures:
https://hg.mozilla.org/integration/mozilla-inbound/rev/ff088809cd2a

Comment 8

[Robert O'Callahan (:roc) (email my personal email if necessary)]

data:text/html,%3C%21DOCTYPE%20html%3E%3Ctitle%3ELoading%20reftest%20analyzer%20for%20Mozilla-Inbound%20Rev3%20WINNT%205.1%20mozilla-inbound%20opt%20test%20reftest%3C/title%3E%3Clink%20rel%3D%22stylesheet%22%20href%3D%22https%3A//tbpl.mozilla.org/css/style.css%22%3E%3Cbody%3E%3Cp%20class%3D%22loading%22%3ERetrieving%20reftest%20log...%3C/p%3E%3Cscript%20src%3D%22https%3A//tbpl.mozilla.org/js/jquery.js%22%3E%3C/script%3E%3Cscript%20src%3D%22https%3A//tbpl.mozilla.org/js/NetUtils.js%22%3E%3C/script%3E%3Cscript%3E%0Afunction%20encode%28s%29%20%7B%20return%20escape%28s%29.replace%28/%3D/g%2C%20%22%253d%22%29%20%7D%0ANetUtils.loadText%28%22https%3A%5C/%5C/tbpl.mozilla.org%5C/php%5C/getLogExcerpt.php%3Fid%3D8012816%26type%3Dreftest%22%2C%0A%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20function%28log%29%20%7B%20window.location.replace%28%22https%3A//hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml%23log%3D%22%20+%20encode%28encode%28log%29%29%29%20%7D%2C%0A%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20function%28%29%20%7B%20%24%28%22p%22%29.removeClass%28%22loading%22%29.text%28%22Fetching%20reftest%20log%20failed.%22%29%20%7D%2C%0A%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20function%28%29%20%7B%20%24%28%22p%22%29.removeClass%28%22loading%22%29.text%28%22Fetching%20reftest%20log%20timed%20out.%22%29%20%7D%29%3B%0A%3C/script%3E

REFTEST TEST-UNEXPECTED-FAIL | file:///c:/talos-slave/test/build/reftest/tests/layout/reftests/scrolling/iframe-border-radius.html | image comparison (==) 

The test failure is caused by the reference stroking all three characters at once, while the test strokes the characters individually. This gives different rendering when the strokes on the characters partially overlap. I'm not sure why this patch changed the behavior here, but we can probably work around it by putting spaces between the characters in the tests so they don't overlap.

Comment 9

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Argh, posted wrong URL

Comment 10

[Robert O'Callahan (:roc) (email my personal email if necessary)]

My guess at the problem was also wrong.

My next guess is that Quartz and D2D are rasterizing path("ABC") differently from path("A") U path("B") U path("C") even where the glyphs don't overlap.

Comment 11

[Edwin Flores [inactive from 2016-12-01] [:eflores] [:edwin]]

Attachment: Patch #3

Just updates the reftest. Draws each letter separately to get rid of weird (negligible) artefacts from drawing more than one glyph at a time.

Comment 12

[Edwin Flores [inactive from 2016-12-01] [:eflores] [:edwin]]

Hmm, ignore that last patch

Comment 13

[Edwin Flores [inactive from 2016-12-01] [:eflores] [:edwin]]

Attachment: Patch Numero Quattro

Again, just updated the failing reftest.

Comment 14

[Edwin Flores [inactive from 2016-12-01] [:eflores] [:edwin]]

Attachment: Same patch, just rebased

Comment 15

[Dão Gottwald [::dao]]

http://hg.mozilla.org/integration/mozilla-inbound/rev/eba8a5ac183f

Comment 16

[Marco Bonardo [::mak] (Away 22 -28 Apr)]

https://hg.mozilla.org/mozilla-central/rev/eba8a5ac183f

Comment 17

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Comment on attachment 589664 [details] [diff] [review]
Same patch, just rebased

>--- a/content/canvas/src/nsCanvasRenderingContext2D.cpp
>+++ b/content/canvas/src/nsCanvasRenderingContext2D.cpp
>+    bool doUseIntermediateSurface = NeedToUseIntermediateSurface()
>+        || NeedIntermediateSurfaceToHandleGlobalAlpha(style);

Note: we put || at the end of the line, not at the start.