The very last vertex of a figure can only be a of type LINE
(non-coincident last/first vertices) or type END (otherwise).
In case the current vertex starts a Bézier path, there is always
at least one more vertex, i.e. (vertex_idx + 1) < vertex_count
is always true.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
When the last vertex is coincident with the first vertex, the last
segment should be suppressed for both END_OPEN and END_CLOSED.
Only when last and first vertex are not coincident the additional
line segment may be added - always for intersection tests and
similar, and for stroking operations when the figure is CLOSED.
Trying to use an zero-length segment in d2d_geometry_intersect_self()
will create invalid segments, causing infinite loops later.
Instead of reducing the vertex_count for coincident first/last
vertices add a dedicated type. This is required as some operations
need the last segment, others do not.
This also allows to remove some replicated code in
StrokeContains()/GetBounds()/Simplify(), as a last Bézier segment
is always processed in the regular loop.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=51139
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This is a preparative patch for the next change, behaviour is
essentially unchanged, though it may be slightly faster.
Rearrange code for outline segment and join generation so it
will work for END_OPEN/END_CLOSED path, with coincident and disparate
last/first vertex.
Each vertex is now fetched once, and pivoted on the next iteration.
Also move invariants in front of the loop,
Path segments are drawn starting with the first segment, up
to vertex_count - 2 (index of start vertex). Only in case of
a END_CLOSED figure with non-coincident last/first vertex,
also the last line segment is drawn.
Joins are added between all drawn segments, and only for
END_CLOSED also the join at the first vertex is added.
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Some software (Buhl Tax and variants) repeatedly calls SetSegmentFlags()
with D2D1_PATH_SEGMENT_NONE, which is just the default value and has no
effect (unless the flags where already changed, which still reports
a FIXME).
Signed-off-by: Stefan Brüns <stefan.bruens@rwth-aachen.de>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
As the test shows, we can create geometries that have less than two vertices
after eliminating duplicates. Calling d2d_cdt_triangulate() on those would
lead to infinite recursion. In principle we could now get rid of the original
vertex count check, but it seems cheap enough that it's worth keeping in order
to avoid some unnecessary work in the somewhat more common case that we have
less than three vertices before duplicate removal.
Based on a patch by Changsheng Chen.
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Currently the ellipse is approximated with quadratic Bézier curves.
Signed-off-by: Giovanni Mascellani <gio@debian.org>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Currently rounded parts are approximated with quadratic Bézier curves.
Signed-off-by: Giovanni Mascellani <gio@debian.org>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
When a rectangle side has zero length, the miter join generator is
confused because it cannot recover the join angle. This patch avoids
that by using artificially non-degenerate previous and next vertices.
Signed-off-by: Giovanni Mascellani <gio@debian.org>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Calling d2d_outline_vertex_set(x, y, 0, 0, 0, 0) causes a normalize() to
be invoked with a zero vector in the vertex shader
vs_code_bezier_outline. This gives rise to an undefinite result, that in
some cases causes the join triangles to be dropped.
Tested on Debian unstable.
Signed-off-by: Giovanni Mascellani <g.mascellani@gmail.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
Stefan Brüns