The code is passing in the extra info received from
the target surface's acquire_source() method to
_cairo_surface_get_extents() rather than the acquired
source surface itself.
Whenever subsurface extents are not contained in the target extents,
using the source image given by the target (with origin corrected by
using an appropriate offset in the data pointer) is not a valid
operation. Fallback to cloning in that case.
Do not try to hide being a subsurface.
This would cause confusion when users where calling surface-type
specific getters on subsurfaces.
FIXME: We still need public API to access a subsurface's target though.
We initialise the context to a non-error state upon creation, but after
destroy there is a window of opportunity where the object is kept alive
inside the context pool and the user could mistakenly keep on passing
the zombie context into cairo functions. As all entry points need to
check error status, flagging the context as an error object upon the
final unreference prevents such misuse (until such as time as the
context is reallocated).
Make linear-gradient-large ref image match xlib gradients (both on nvidia
and on macosx) and pixman gradients (when patched to compute the linear
gradients using floating point math).
Insertion sort can take up to O(n^2), mergesort is guaranteed to run
in O(n*log(n)).
An example showing bad performance for insertion sort is:
https://bugs.freedesktop.org/show_bug.cgi?id=28067
The mergesort has been engineered to be fast even when working on
cases where the insertion sort would have performed well and as
expected it shows no changes in the benchmark cairo traces.
Vertical RGB mapping previously forced opaque pixels.
To be consistent with horizontal RGB/BGR and vertical BGR it
should use an alpha equal to the mid channel (green).
Add NONPKGCONFIG_EXTRA_LIBS that are only put into the generated pc file
but not used when linking. This is now used to add -lcairo-gobject to
the cairo-gobject.pc file.
Make the loops count depend on the actual calibration_loops/calibration_time
instead of calibration_loops/calibration_max_time.
This avoids having some tests take much less/more than the wanted time per iteration
(I was having some tests taking about 1 second, other taking about 7 seconds when
the ms_per_iteration was 2000)
Spend 0.5-1 times the time wanted for each iteration in calibration, increase the
accuracy of loops count. Just making the loops count be the correct ratio doesn't
guarantee that the iteration time is accurate. By actually measuring iteration
times until it gets greater than 1/4 of the wanted time, the total sum is bound
to be <= the wanted iteration time and last calibration time is between 1/4 and
1/2 of the wanted time, so it should give a very accurate loop count.
The bounds checks on the rectangle to blit were done
after converting target/source to inclusive/noninclusive
coordinates rather than before, potentially causing an
off-by-one in the sizes, since the same logic was applied
to both inclusive and non-inclusive coordinates. This
patch enforces bounds up front.
Thread "OS/2 surface fixes" on cairo-l:
http://lists.cairographics.org/archives/cairo/2010-July/020343.html
Use a single code path to release the resources acquired
for a surface that's been partially constructed, rather than
have multiple error exits which each release the resources
acquired so far.
Thread "OS/2 surface fixes" on cairo-l:
http://lists.cairographics.org/archives/cairo/2010-July/020343.html
Clarify the documentation for cairo_os2_surface_create()
and cairo_os2_surface_set_hwnd() to note that the ownership
of the presentation space and window respectively
remains with the caller.
Thread "OS/2 surface fixes" on cairo-l:
http://lists.cairographics.org/archives/cairo/2010-July/020343.html
On arithmetic overflow the buffer allocator would attempt
to allocate zero bytes instead of failing the allocation.
Depending on the implementation of the underlying allocator
being proxied, this may result in an allocation of a zero
length buffer, causing all kinds of grief. This patch
causes arithmetic overflows to explicitly fail the allocation.
Thread "OS/2 surface fixes" on cairo-l:
http://lists.cairographics.org/archives/cairo/2010-July/020343.html
There's a format mismatch between cairo's pixel formats
and OS/2 24 bpp data: the individual pixels in OS/2 take
only three bytes per pixel, but there needs to be padding
between lines to make the start of a line always be aligned
at a DWORD boundary.
This patch fixes the format conversion code to include
the required padding between rows.
Thread "OS/2 surface fixes" on cairo-l:
http://lists.cairographics.org/archives/cairo/2010-July/020343.html
Degenerate linear patterns are considered clear if they have
EXTEND_NONE, the average of the first and the last stop if they are
EXTEND_PAD, the weighted average of the stops (based on the size of
the interpolation range in which they are active, just like integrating
over the whole interpolation range and taking the average) if they are
EXTEND_REPEAT or EXTEND_REFLECT.
Fixes degenerate-linear-gradient
Introduce two new (private) functions for testing the degeneracy of linear
and radial gradients and use them instead of verbose (and error-prone) tests.
Two clear stops with different "underlying" colors must be considered
different, since when interpolated with non-clear colors they will
produce different results.
Quartz previously crashed with NULL backends and didn't check for
the backend type when getting a CGContext from a quartz surface,
returning meaningless data for subsurfaces.
Falling back when painting would ignore the alpha value (which is
needed to have the correct mask opacity).
_cairo_quartz_surface_paint_cg doesn't fallback, so the usual mask
fallback path is now taken, as expected.
Self-intersecting strokes were drawn incorrectly when an unbounded
operator was used, since the fixup operation also cleared the
intersection.
Fixes clip-stroke-unbounded.
Merge clip-*-unbounded tests and add self-intersections to the paths
that are drawn. This exposes a bug in the unbounded fixup code in quartz.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The interpolation range of repeating radial gradients can safely be
reflected around any integer (previously 0), but for reflect-extended
radial gradients can only be reflected around odd integers if the
appearance is to be the same, thus reflecting around 1 is correct for both.
Fixes radial-gradient.
