296463 - update gfx/cairo to cairo 0.5

Status: RESOLVED FIXED

(Core Graveyard :: GFX, defect)

Description

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Three separate patches here, based on roc's work moving things to cairo 0.5.  I
only have the changes to gfx/cairo and to canvas and svg here -- I figure roc
can land his gfx/src/cairo work separately, since it's not part of the build.

Comment 1

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Attachment: cairo-update.patch

This patch brings gfx/cairo up to 0.5.0/pixman 0.1.5.

Comment 2

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Attachment: canvas-update.patch

This patch brings canvas up to cairo 0.5.0.  There's a significant performance
improvement with this patch (like 3-4x), and I have followup patches later that
will improve performance even more.

Comment 3

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Attachment: svg-update.patch

SVG update for cairo 0.5.0, based on roc's patches.

Comment 4

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

Comment on attachment 185225 [details] [diff] [review]
cairo-update.patch

Looks great.

I have additional patches to cairo to fix a clipping bug, but we can land that
later.

Comment 5

[tor]

Something odd is happening with the svg patch - some strokes are vanishing.

Comment 6

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

(In reply to comment #5)
> Something odd is happening with the svg patch - some strokes are vanishing.

Hmm, take a look at the dash-dos.diff -- I forgot to mention that I had to
comment out the alpha modulation if dashes are invisible, since alpha is no
longer stored as a gstate member.   I don't /think/ this is causing the stroke
disappearing, but I'm not positive.

Comment 7

[tor]

Attachment: additional svg patch

cairo's save/restore behavior with respect to paths changed.  This
additional patch atop the other svg one is needed.

Comment 8

[tor]

Comment on attachment 185227 [details] [diff] [review]
svg-update.patch

>--- layout/xul/base/src/Makefile.in	monotone
>+++ layout/xul/base/src/Makefile.in	monotone
>@@ -150,7 +150,7 @@
> 
> DEFINES += -D_IMPL_NS_LAYOUT
> 
>-ifdef MOZ_CAIRO
>+ifdef MOZ_ENABLE_CAIRO
> CFLAGS		+= $(MOZ_CAIRO_CFLAGS)
> CXXFLAGS	+= $(MOZ_CAIRO_CFLAGS)
> endif

If this bit of the patch is still needed, that should probably read
MOZ_ENABLE_CANVAS.

With that and the attached additional patch for save/restore, r=tor.

Comment 9

[tor]

Comment on attachment 185227 [details] [diff] [review]
svg-update.patch

SVG images and group opacity are also broken (surface code).

Comment 10

[tor]

Also problems with text positioning on partial redraws (drag another window
across  mozilla displaying svg text).

Comment 11

[tor]

SVG images and group opacity can be fixed by passing 0,0 instead of width,height
to the two calls to cairo_set_source_surface() in nsSVGCairoCanvas.cpp.

Comment 12

[tor]

Text positioning problem can be fixed by replacing the cairo_transform() in
nsSVGCairoGlyphGeometry::GetGlobalTransfrom() with cairo_set_matrix().

Comment 13

[tor]

Group opacity broken for partial updates.  Test url:

http://www.w3.org/Graphics/SVG/Test/20030813/htmlframe/full-masking-opacity-01-b.html

Comment 14

[tor]

Attachment: replacement svg patch (includes all fixes plus group opacity partial update fix)

Comment 15

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

I'll land all this late tonight, when I can have a quiet half hour or so with
the tinderbox.  I'll probably land cairo 0.5.0 first and tag gfx/cairo, and then
apply our patches on top as a separate commit.

Comment 16

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Checked in, albeit grudgingly on OSX and Win32 -- both of them had crazy
#include issues, and I don't have an OSX build environment set up and my win32
desktop died yesterday.  I bandaided the issues, but I'll try to figure out
what's really happening in the next week or so.  (The bandaids shouldn't affect
any functionality, they just shouldn't be needed.)

Comment 17

[(not currently active) Ted Mielczarek]

Pretty sure this broke drawWindow, at least on Win32, see bug 296697