Do this by tiling the surface form the solid pattern. The pattern in
turn calls into xlib's create_solid_surface which returns a dithered
pattern.
This can get into infinite loop right now, because of the way solid
surface cache tries to repaint cached surfaces.
Otherwise we can't do dithering. This drastically improves gradient rendering
on 16bit displays, essentially making them indistinguishable from 32bit ones
with a naked eye.
Remove the intermediate rgb333 for PseudoColor and work on the
cube directly. Also upgrade to a 6x6x6 cube instead of 5x5x5.
Do dithering on both PseudoColor and TrueColor, using a 4x4 pattern.
This only affects X servers with no XRender.
This can be used by _cairo_pattern_acquire_surface_for_solid() to let
the backend create custom surfaces for solid fill instead of a 1x1
similar surface.
This was added in 41c6eebcd1, to avoid
allocating short-lived patterns. However, this was error prune as the
color information was duplicated in the pattern and could get out of
sync. Indeed, it was out of sync before this commit in the call from
cairo-clip.c.
By allocating the solid pattern on the stack we fix the original problem
without creating new ones.
The code in that cache is becoming uglier and uglier. But at least it
works correctly again. Time for a rewrite.
(Also move the destroy after unlocking)
Reported on the cairo mailing list:
http://lists.cairographics.org/archives/cairo/2008-May/014233.html
The tessellator would sometimes produce self-intersecting trapezoids
because it would skip the last edge in the active list when deciding
whether we can continue the current trapezoid or not.
The bug never caused a problem with pixman based rasterisation
because pixman stops filling in a trapezoid once it detects a self
intersection.
Replace a clip rectangle that covers the entire surface with no
clipping as this is quite a common operation when toolkits (i.e. GTK+)
clear surfaces before drawing and avoids a redundant
XRenderSetClipRectangles.
It is possible for an XRender capable surface to use
_cairo_xlib_surface_solid_fill_rectangle() if the surface
HAS_CREATE_PICTURE() && ! HAS_FILL_RECTANGLES(), in which case we need to
handle the surface having no associated visual.
Fixes test/xlib-expose-event.
It appears that the PowerPC Linux ABI specifies a number of magic
functions for register saves and restores which currently appear as
exported symbols in the shared object. Ignore them.
Instead of allocating the union of all possible pattern types, just
allocate the specific pattern as used by the function in order to trim
the stack space consumption and flag potential misuse.
Behdad Esfahbod complained that commit
8457374c9c overwhelmed the function with
added complexity and arbitrary limited a backend to one-quarter of the
cache. The elegant solution, he outlined, was to look at the surface that
would be evicted and if possible repaint it, instead of creating a
replacement. This not only simplifies the code, reduces the number of
checks performed to find a match (or victim) and allows the cache to be
naturally shared between the various backends.
In order to avoid re-rasterising a glyph that is pending an
XRenderFreeGlyph, we first scan all glyphsets and their arrays of
pending_free_glyphs for a matching glyph. The additional cost of
scanning the extra arrays should be negligble as most fonts will only
have the single array (which we would scan anyway) but we potentially
save an expensive rasterisation and short-lived image surface.
(As suggested by Behdad Esfahbod.)
Enlarge the embedded arrays of edges and points for cairo_polygon_t and
cairo_spline_t respectively, such that the frequent allocations are
eliminated whilst running a firefox3 benchmark.
First try to allocate the vertices using an on-stack array, otherwise, if
we need more vertices than can be accomodated, fallback to using a heap
array.
That is, instead of uniformly spreading the colors at 0/4, 1/4, 2/4, 3/4, 4/4
intensities, we do 0/7, 1.5/7, 3.5/7, 5.5/7, 7/7 now. Those better
approximate the 0/7..7/7 of the rgb333 space that we first convert to.
