We must destroy glyph cache surface in device_finish instead of in
device_destroy because in device_destroy device status is
DEVICE_FINISHED and the operation is invalid.
A few test cases purposely create fractional surface sizes which can not
be natively supported by the raster backends such as GL. For these
backends we need to consistent in creating a surface that is large
enough to contain the test, so we need to use ceil() rather than
implicit truncation to integers.
A consequence of the misalignment between the Window size and the
surface size (where one was using ceil and the other not) is that the
first row of the cairo surface would not be visible on the output.
Based on a patch by Chuanbo Wen to fix 5 test cases, such as
group-unaligned.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The size of the target area doesn't really have much to do with the size of the
recording surface that we are painting from. Thus, let's use the recording
surface's size instead.
Since we apply the transformation before replaying the recording surface, we
need to transform the recording surface's size via the inverse of our pattern
matrix to get the size in the target surface. This makes this a little more
complex.
Signed-off-by: Uli Schlachter <psychon@znc.in>
Let's say we are painting recording surface 'source' to xcb surface 'target' by
replaying the source to a temporary surface 'tmp'.
Previously, the xcb backend replayed the recording surface to tmp with just a
translation and then used that as its source surface with the pattern's
transformation. That means 'tmp' used the same coordinate system as 'source'.
This patch changes this so that the transformation is applied during the replay
and painting from 'tmp' to 'target' is just a simple translation, so 'tmp' now
uses the same coordinate system as 'target'.
This should produce way less better results, because transforming a recording
surface should have less artifacts than transforming a raster surface.
Fixes: record1414x-* record2x-* record90-* ps-surface-source
Breaks (or rather, "exposes unrelated bug that I have not yet figured out in"):
record-extend-*-similar
Signed-off-by: Uli Schlachter <psychon@znc.in>
When the desintation surface is not a texture, it is flipped in the Y
axis. So we need to correct the Y coordinates when using glScissor to
the set the clip region.
Fixes 14 cases in cairo-test-suite, for example partial-clip-text-top
If the angle between two segments is small we can simply replace the
round-join with a bevel-join. This is done automatically by the
insertion of the triangle fan as it will not be able to find a point
around the pen between the two vectors. However, we can make that search
cheaper by inspecting whether the bisection angle is small enough that
the bevel-join will be within geometric tolerance of the round-join.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Near an inflection, the angle between two segments of a spline increases
rapidly (as the radius of curvature decreases for the cusp). The angle
may increase so much that a simple line connecting the two outside
points of the spline is not within the user specified geometric
tolerance (with the result that you can generate severe ugliness around
a cusp). Extend the current detection of the exact inflection to cover
the sharp joins near the cusp by inspecting whether the bisection angle
is larger than acceptable.
Fixes bug-spline.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Carl Worth demonstrated a glaring bug in the new stroking code,
introduced in commit 545f30856a (stroke: Convert the outlines
into contour and then into a polygon), whereby only a bevel join was
being used to connect segments around a sharp inflection point.
This adds the two examples he reported to the test suite.
In the case we try to use an unbounded operation, passing a NULL clip
causes that operation to clear the rest of the surface. Instead we need
to trim the _cairo_surface_mask() to the operation extents.
Fixes overlapping-glyphs.
Suggested-by: Chuanbo Weng <strgnm@gmail.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Jose Dapena Paz reported an assertion following the uninitialised status
value being returned. Also the function failed to free its allocations.
Based on a patch by Jose Dapena Paz <jdapena@igalia.com>.
Reported-by: Jose Dapena Paz <jdapena@igalia.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=51104
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
_cairo_surface_subsurface_set_snapshot () sets the subsurface as the
snapshot of its target. This creates a reference cycle (as the target
is already referenced by the surface) and thus a memory leak (assuming
the likely case that user doesn't call finish).
Test case: subsurface-similar-repeat.
So make this call as a no-op for the time being until the bug is fixed.
quartz-image uses _cairo_surface_is_image(), which is now declared in
cairo-image-surface-inline.h.
Fixes:
cairo-quartz-image-surface.c: In function 'cairo_quartz_image_surface_create':
cairo-quartz-image-surface.c:312: error: implicit declaration of function '_cairo_surface_is_image'
cairo-quartz-image-surface.c:312: warning: nested extern declaration of '_cairo_surface_is_image'
This new pixman API allows glyphs to be cached and composited in one
go, which reduces overhead compared to individual calls to
pixman_image_composite_region32().
Notes:
- There is an explicit call to _cairo_image_scaled_glyph_fini(). This
could instead be done with a private, but I chose not to do that
since we don't need to store any actual data; we only need
notification when the glyph dies.
- The slowdown in poppler-reseau is real and stable across runs. I'm
not too concerned about it because this benchmark is only one run
and so it is dominated by glyph cache setup costs and FreeType
rasterizing.
Performance results, image backend:
Speedups
firefox-talos-gfx 5571.55 -> 4265.57: 1.31x speedup
gnome-terminal-vim 1875.82 -> 1715.14: 1.09x speedup
evolution 1128.24 -> 1047.68: 1.08x speedup
xfce4-terminal-a1 1364.38 -> 1277.48: 1.07x speedup
Slowdowns
poppler-reseau 374.42 -> 394.29: 1.05x slowdown
Performance results, image16 backend:
Speedups
firefox-talos-gfx 5387.25 -> 4065.39: 1.33x speedup
gnome-terminal-vim 2116.66 -> 1962.79: 1.08x speedup
evolution 987.50 -> 924.27: 1.07x speedup
xfce4-terminal-a1 1856.85 -> 1748.25: 1.06x speedup
gvim 1484.07 -> 1398.75: 1.06x speedup
Slowdowns
poppler-reseau 371.37 -> 393.99: 1.06x slowdown
Also bump pixman requirement to 0.27.1.
This is hopefully a lesser used path and the attempted optimisation to
continue a stopped edge with a colinear stopped edge highly unlikely and
lost in the noise of the general inefficiency of the routine. As it was
broken, rather than attempt to rectify the "optimisation" remove it.
Reported-by: Evangelos Foutras <evangelos@foutrelis.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50852
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As it turns out due to the rules of polygon intersection, there is never
any overlapping spans so the choice is arbitrary. However, lets be
consistent with the rest of the code.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If the font metrics appear broken, i.e. key values are being reported as
zero, skip approximating the bbox of the glyph string.
References: https://bugs.freedesktop.org/show_bug.cgi?id=50688
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
In addition to fixing a bug 7d8d98b91c releated to
expanding a8 glyphs into a8r8g8b8, this commit also added an
optimization where if the first glyph had format a8r8g8b8, the mask
was created in this format from the beginning instead of later
converting from a8 to a8r8g8b8.
However, the optimization had two bugs in it:
(1) The computed stride was 3 * width, not 4 * times width, and
(2) In the case where the mask was allocated on the stack, it was
allocated as PIXMAN_a8 and not a8r8g8b8.
The commit fixes both bugs.
The check for standalone headers requires that the
cairo-image-surface-inline.h include the cairo-surface-private.h
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Coordinate translation is necessary in composite_boxes(), composite()
and composite_traps(). The translation value should set to correct
value in traps_to_operand(). Also fix the bug in lerp().
We need to scale the channels of the glyph into the destination (and
indeed expand a8 into a8r8g8b8) when adding into the mask. Normally we
have matching formats for the glyph surfaces and the temporary mask,
for which we can continue to take the faster path.
Reported-by: Søren Sandmann <sandmann@cs.au.dk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
An overzealous update after converting antialiasing missed the object of
this test was exactly to point out an error due to the antialiasing. So
restore it back to the prestine reference and mark the image backend as
failing.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Every time we run ./autogen.sh, the INSTALL file would be replace by
/usr/share/automake-1.11/INSTALL, which changes the file. This is very
annoying since it messes up the result of 'git diff' and git always
wants to commit it.
Declare it as foreign would fix this annoying issue. Since it's not a
GNU project, it makes sense to do so.
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Use _cairo_gl_operand_copy to do the operand copy instead of by
something like *operand = surface->operand.
This would fix a crash introduced by operator-source test case, which
forgot to do reference while copying operand but did dereference while
destroying surface at combine_clip_as_traps().
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
[ickle: drop the extra reference for the owned surface]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
and reuse it to implement acquire_source.
Fixes
cairo-quartz-surface.c: In function '_cairo_quartz_surface_map_to_image':
cairo-quartz-surface.c:1568: warning: return from incompatible pointer type
cairo-quartz-surface.c:1576: warning: return from incompatible pointer type
