The PDF surface was adding extra stops at the 0.0 and 1.0 offset when
there was not already stops at these offsets. This has been replaced
with code to move the coordinates of the linear gradient line in to
the position of the first and last offset.
If scaled_font_destroy() is called, a deadlock can result; there's no
reason to call destroy since the initialization failed (and, indeed,
it might not be valid to do so anyway).
Patch from: Robert O'Callahan <roc@ocallahans.org>
Add cairo_win32_font_face_create_for_logfontw_hfont, allow win32
scaled_fonts to rescale themselves properly to the required CTM and
only use the font_face's hfont if we're sure it's appropriate.
The following 7 tests currently fail with poppler due to:
Poppler does not correctly handle knockout groups
https://bugs.freedesktop.org/show_bug.cgi?id=12185
and we've verified with acroread that the cairo-pdf output
does render as intended there. The disabled tests are
clip-operator, operator-clear, operator-source, over-above-source,
over-around-source, over-below-source, and over-between-source.
The stream handling has been changed to support writing the content to
one or more group objects. Each page has a top level knockout
group. The first operation in the knockout group paints another group
containing the content. Fallback images are painted from the knockout
group. This ensures that fallback images do not composite with any
content under the image.
These are useful programs, but they don't belong in test/.
They are never used in the test suite at all. Instead, these
should exist in some demo package, or as applications in their
own right.
(The motivation for the removal is that someone just encountered
a build error with one of these. And I'm not interested in debugging
build errors for unused components.)
A pixman bug had kept this from being correct earlier. It
works now as long as pixman includes the following fix:
8ff7213f39edc1b2b8b60d6b0cc5d5f14ca1928d
rasterize traps that extend outside of the image bounds correctly
I've been seeing failures with the PS backend for these tests
for a while. This is in spite of the face that we've had a PS-
specific reference image for each of these tests. I'm not sure
what the difference is between my current system and the
system used to generate the original ps-specific reference
images.
One thing we definitely need to do is to bundle fonts with the
test suite to avoid problems like this. Another is to start
documenting the systems used to generate reference images.
I get very confused with ghostscript versioning, but I can
at least say that "gs --version" reports 8.15.3 here, (though
the package appears to have an 8.56 version).
The following 4 tests recently started drawing with native operations
in the place of image fallbacks, so we need ps-specific reference images
to account for minor rasterization differences: over-above-source,
over-around-source, over-below-source, and trap-clip.
This method is for use in vector backends, where fill immediatly followed by
stroke command with the same path can be emited in the same backend command.
This commit also factorize the detection of such cases in the meta surface
backend and automatically call the fill_stroke method on replay.
Whilst testing the fallback surface, the resultant image was being
clipped to the screen size. Be conservative and refuse to create
windows (for CAIRO_CONTENT_COLOR surfaces) that are larger than the
screen.
As well as marking the XRender extension unavailable in
_cairo_boiler_player_xlib_disable_render(), we need to clear any
XRender derived information stored during the surface creation.
These were recently added, (as part of sparse integration?), but they
break boilerplate which reaches into at least cairo-types-private.h
and cairo-scaled-font-private.h. But boilerplate cannot see cairoint.h
or else it gets the internal sybol renaming, (with the INT_ prefix),
and then all the test suite tests refuse to link.
If this change reverts some recently-added functionality, (or
cleanliness), then we'll just need to find some other way to add that
back again without the breakage.
Apparently, our much-beloved error message, (just look at how many people
love to quote it in bugzilla entries), was removed during the switch to
the new pixman. It was replaced with an obscure failure of a nil surface
being returned, (leading to obscure failures or false "out of memory"
errors).
So we're putting this back now, (even though we _really_ want to ensure
that this message never gets printed in practice).
Previously, the code was just using cairo_format_t which is much more limited
than the formats supported by pixman, (so many "odd" X server visuals would
just fall over).
