ARM has much weaker memory ordering guarantees than x86 by default, and
needs the memory barriers. A similar fix exists in the Skia codebase.
Fix suggested by Nathan Froyd.
Fixes: FDO #90302
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Acked-by: "Henry (Yu) Song" <henry.song@samsung.com>
This function tries to use _cairo_xcb_connection_put_image() to do the actual
work. However, that function can only be used for images with "native" stride.
If we only want to upload a rectangle from within an image, the function
_cairo_xcb_connection_put_subimage() has to be used. This function makes sure
that the correct information is sent to the X11 server.
No unit test for this, because we currently do not test the !SHM case
automatically. Perhaps we should?
Signed-off-by: Uli Schlachter <psychon@znc.in>
This function always returned CAIRO_INT_STATUS_SUCCESS, even if it didn't do
anything. This commit makes the function return CAIRO_INT_STATUS_UNSUPPORTED
instead.
No unit test for this, because we currently do not test the !SHM case
automatically. Perhaps we should?
Signed-off-by: Uli Schlachter <psychon@znc.in>
Some broken pdfs use glyph 0 in embedded fonts for rendering instead of .notdef.
The cmap we use for embedding latin fonts does not allow rendering glyph 0. Ensure
if glyph 0 is used, it is mapped to a non 0 glyph in the subset.
Bug 89082
The memory allocated to "image" at line 298 is not freed before moving to label fail at line 305 and 314.
This patch takes care of this memory leak in above mentioned cases.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=91537
Signed-off-by: Sahil Vij <sahil.vij@samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
The attribute list is terminated by GLX_NONE (defined as 0x8000), but
the man page of 'glXChooseVisual' says it must be terminated with None
(0L).
Issue found and fix suggested by Massimo.
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=91293
In read_file(), we see:
*data = malloc (*len);
...
if (fread(*data, *len, 1, fp) != 1) {
free(data);
...
The free call needs to be free(*data), to match the malloc call.
Matthias Clasen found this via Coverity and proposed the fix.
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=91381
returned traps will have their left edge to the left of their right
edge, but if only one trap is passed in then the function always returns
without doing anything. This results in incorrect rendering of SVG
paths with more than one subpath.
Currently calls to _cairo_bentley_ottmann_tessellate_rectangular_traps
are guarded by traps.has_intersections checks, so this is only a
theoretical bug. But we'll eliminate the potential of the bug by
making the left side to be left of the right side, similar to what was
done in _cairo_bentley_ottmann_tessellate_boxes (commit 11b6c49c).
Patch authored by Tom Klein for Mozilla.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=90984
Ref: https://bugzilla.mozilla.org/show_bug.cgi?id=853889
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
This reverts commit c8b6c6066a.
An undefined macro is evaluated as 0.
The preference is to use #if MACRO_NAME rather than #ifdef.
Conflicts:
src/cairo-gl.h
Some blend operators have a (small, but potentially noticeable)
different behaviour in Quartz and pixman. Use the Quartz ones only for
the operators which match pixman except for rounding errors.
Fixes:
- extended-blend
- extended-blend-alpha
- extended-blend-mask
- extended-blend-alpha-mask
- extended-blend-solid
- extended-blend-solid-alpha
The Quartz framework provides several quality settings, but they were
not used. Now the translation of cairo_filter_t to Quartz filtering
modes tries to match the quality settings from image as much as
possible.
Specifically, CAIRO_FILTER_GOOD and CAIRO_FILTER_BILINEAR are now
converted to kCGInterpolationLow, which seems to be a bilinear filter.
Fixes:
- pixman-downscale-bilinear-24
- pixman-downscale-bilinear-95
- pixman-downscale-good-24
- pixman-downscale-good-95
- pthread-same-source
- recording-surface-extend-none
- recording-surface-extend-reflect
- recording-surface-extend-repeat
- recording-surface-over
- recording-surface-source
- surface-pattern-big-scale-down
- surface-pattern-scale-down
- surface-pattern-scale-down-extend-none
- surface-pattern-scale-down-extend-reflect
- surface-pattern-scale-down-extend-repeat
This makes the results of the test suite more stable across different
environments, because it does not rely anymore on
CAIRO_FONT_FAMILY_DEFAULT (which on Windows is "Arial", on Mac
"Helvetica").
This change should not affect Linux environments, assuming that the
default font is already set to "DejaVu Sans".
The sizeof operator now applied to the correct variable "ctx->glyph_cache",
instead of its pointer address, in function "_cairo_gl_composite_flush".
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=91321
Signed-off-by: Arpit Jain <jain.arpit@samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
The pattern structure is hardcoded to hold 7 elements, yet the
pattern_names array in cairo-surface-observer.c is initialized with 8
strings. This causes a crash in print_array at line 1587 when it tries
to access the 8th member.
Hence changed the 'type' array from type[7] to type[8] to avoid out of
bound access.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=91266
Signed-off-by: Ashim <ashim.shah@samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
The pointer 'filename' is already freed and still used as a function
argument. This patch will free the pointer 'filename' only after it is
used.
Also, the patch ensures that it frees the pointer 'filename' before any
return of this function.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=91206
Signed-off-by: Arpit Jain <jain.arpit@samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
In commit f6843d5cbb
Author: Arpit Jain <jain.arpit@samsung.com>
Date: Mon Jul 6 14:13:06 2015 -0700
xlib: Fix deferencing of uninitialised 'display'
the common error + clenaup path was clumsily fixed to use the right
variable after the error didn't set the local display variable.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Initialising 'display' to NULL and checking before deferencing during display->base.
This patch will check the deferencing of uninitialised 'display' in case,
_cairo_xlib_display_acquire does not return CAIRO_STATUS_SUCCESS.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=87893
Signed-off-by: Arpit Jain <jain.arpit@samsung.com>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
There is an unintentional (benign) missing line continuation "\" in the
definition of MAYBE_WARN in configure. configure completes and issues a
WARNING:
configure: WARNING: cache variable cairo_cv_warn_maybe contains a
newline
Found and solved by Larry Baker.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=89750
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
When you draw (part of) a surface to an XCB surface, the XCB backend will safe
the uploaded part as a snapshot to the input surface. This allows to re-use this
picture in case the same surface is later used again as a source.
However, other backends do not do this and this has caused and/or highlighted
numerous bugs. Just skipping the snapshot fixes or hides these bugs.
Papers-over: https://bugs.freedesktop.org/show_bug.cgi?id=67505
Signed-off-by: Uli Schlachter <psychon@znc.in>
That is if the difference between the origin and the end is bigger than
.5 round up regardless of the coordinates.
Round the difference of the floats instead of rounding the floats then
diff them.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=84396
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
On AIX, the token jmpbuf is a pre-processor macro.
cairo-script-scanner.c includes a private struct with a member named
jmpbuf which gets renamed to __jmpbuf when AIX's sys/context.h has been
included.
While judicious ordering of includes might kludge around this problem
(by causing all references to .jmpbuf to become .__jmpbuf), it's better
to simply select a new name for the struct member that won't suffer the
collision.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=89339
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Use quoted includes rather than bracketed, to prefer linking to the
in-tree cairo in preference to the system cairo.
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Chris wrote all of the cairo script stuff. I'm making a guess about the
copyright date.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
den_det is positive because intersect_lines is called
only after _slope_compare returned > 0 and slope_compare
is returning the sign of den_det
The quadratic-time intersection finder is #if 0-ed out
in src/cairo-bentley-ottman.c, but is unusable even there
since the second commit to that file.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=74779
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
Optimizing compilers aggressively remove code that is executed only
after an undefined behaviour occurred.
Also, the difference of two (non char) pointers hides an integer
division that, because the divisor is known at compile time, is
transformed into a multiplication by a pseudo-reciprocal, and in this
case the difference is not always a multiple of the divisor, resulting
in an invalid comparison predicate.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=74779
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
