Commit 337ab1f8d9 introduced delayed
font destruction to overcome a race, but prevented the correct cleanup
of the font cache.
This caused fonts to stay in the cache and caused a crash in the
api-special-cases (when running the test suite with -f).
Fixes api-special-cases,
https://bugs.freedesktop.org/show_bug.cgi?id=87567
The snapshot takes a reference to the target recording surface in order
to enable it for use by multiple treads. In order to balance this, the
other two sources of recording surface must also take a reference and
for us to release that reference after the replay.
Otherwise, we end up with a memory leak:
==1== 1,392 bytes in 3 blocks are definitely lost in loss record 1 of 7
==1== at 0x4A06BCF: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==1== by 0x4C7042D: _cairo_recording_surface_snapshot (cairo/src/cairo-recording-surface.c:1427)
==1== by 0x4C842BE: _cairo_surface_snapshot_copy_on_write (cairo/src/cairo-surface-snapshot.c:189)
==1== by 0x4C7E7E0: _cairo_surface_detach_snapshot (cairo/src/cairo-surface.c:348)
==1== by 0x4C7E55B: _cairo_surface_detach_snapshots (cairo/src/cairo-surface.c:333)
==1== by 0x4C7E55B: _cairo_surface_flush (cairo/src/cairo-surface.c:1545)
==1== by 0x4C7E6CC: _cairo_surface_finish_snapshots (cairo/src/cairo-surface.c:1017)
==1== by 0x4C7E6CC: cairo_surface_destroy (cairo/src/cairo-surface.c:961)
==1== by 0x4C625A7: cairo_pattern_destroy (cairo/src/cairo-pattern.c:1131)
==1== by 0x4C3FAC6: _cairo_gstate_fini (cairo/src/cairo-gstate.c:225)
==1== by 0x4C3C68C: _cairo_default_context_fini (cairo/src/cairo-default-context.c:75)
==1== by 0x4C3C708: _cairo_default_context_destroy (cairo/src/cairo-default-context.c:93)
==1== by 0x43E576: record_get (cairo/test/record-extend.c:158)
==1== by 0x43E576: record_replay (cairo/test/record-extend.c:173)
==1== by 0x40E22D: cairo_test_for_target (cairo/test/cairo-test.c:929)
==1== by 0x40E22D: _cairo_test_context_run_for_target (cairo/test/cairo-test.c:1532)
==1== by 0x40B6C0: _cairo_test_runner_draw (cairo/test/cairo-test-runner.c:255)
==1== by 0x40B6C0: main (cairo/test/cairo-test-runner.c:937)
Reported-by: Massimo Valentini <mvalentini@src.gnome.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87898
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
When running the cairo-test-suite, valgrind reports each of the
registered test as a leak, because they are _list_prepend()'ed, but
the tests list is never _list_free()'d.
Fixes the following valgrind error:
malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
cairo_boilerplate_xmalloc (cairo-boilerplate-system.c:47)
cairo_test_register (cairo-test-runner.c:131)
_cairo_test_runner_register_tests (cairo-test-constructors.c:1112)
main (cairo-test-runner.c:714)
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
The pattern created by cairo_pattern_create_rgb() is owned by the
caller, hence it needs to be released.
Fixes the following valgrind error:
malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
_cairo_pattern_create_solid (cairo-pattern.c:605)
_cairo_pattern_create_in_error (cairo-pattern.c:628)
cairo_pop_group (cairo.c:552)
test_cairo_push_group (api-special-cases.c:157)
preamble (api-special-cases.c:1766)
main (cairo-test-runner.c:228)
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
On 32bit SPARC the scan converter was causing a SIGBUS due to an unaligned
memory access while accessing an int64_t. This memory accessing was to struct
quorem's rem member.
This crash occurred because the tor-scan-converter contains its own
implementation of a memory pool. This implementation only guarantees an
alignment of sizeof(void *), which is less than what a 64 bit type requires on
32bit platforms. This 4 byte alignment is guaranteed, because struct _pool_chunk
(which is the struct that is used for managing free space) contains elements of
that size and so the size of that struct is a multiple of this size as well.
This problem was introduced with commit 03c3d4b7c1.
To fix this problem, this commit introduces a int64_t member to struct
_pool_chunk that marks the beginning of the free data space. Thanks to this, the
compiler ensures proper alignment and sizeof(struct _pool_chunk) becomes a
multiple of 8.
However, previously the end of the struct marked the beginning of the data and
sizeof() was used for correctly calculating offsets to the data section. So,
just adding such a member would work, but would also waste some memory. To avoid
this, this commit also changes the rest of the pool implementation to
accommodate.
Reported-by: Nicolas Setton <setton@adacore.com>
Signed-off-by: Uli Schlachter <psychon@znc.in>
Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
If a span length is negative don't go out of bounds processing the fill
data.
Patch thanks to Ilya Sakhnenko <ilia.softway@gmail.com> on mailing list.
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
$ ./check-doc-syntax.sh
Checking documentation for incorrect syntax
./cairo-pattern.c (3342): ERROR: Will invalid doc id (should be 'cairo_...:')
The proposed changes fixes the warning about the documentation style used in cairo.
Signed-off-by: Ravi Nanjundappa <nravi.n@samsung.com>
malloc(0) needn't return NULL, and on glibc, doesn't. Then we encounter
a loop of the form do { ... } while (--c), which doesn't do quite what
you were hoping for when c is initially 0.
Since there's nothing to swap in this case, just bomb out.
Signed-off-by: Adam Jackson <ajax@redhat.com>
malloc(0) needn't return NULL, and on glibc, doesn't. Then we encounter
a loop of the form do { ... } while (--c), which doesn't do quite what
you were hoping for when c is initially 0.
Since there's nothing to swap in this case, just bomb out.
Signed-off-by: Adam Jackson <ajax@redhat.com>
since the PS Document Structing Conventions impose a 255 character
line limit. PDF does not require wrapping.
pdf-operators is designed to emit the same output for PS and PDF.
Unfortunately some PDF interpreters don't like strings split with
'\\\n' and some PS interpreters don't like strings split with ')('.
So we are forced to make pdf-operators handling string wrapping
differently for PDF and PS.
Bug 85662
These two images are mis-rendered (clearly evident from visual
inspection). By removing them, the test will fall back to the more
general format-specific images, huge-radial.argb32.ref.png and
huge-radial.rgb24.ref.png.
Note that the huge-radial.pdf tests still fail to pass, but the pdiff
looks more sensible.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=66218
Signed-off-by: Bryce Harrington <bryce@bryceharrington.org>
$ ./check-doc-syntax.sh
Checking documentation for incorrect syntax
./cairo-pattern.c (3346): ERROR: Will bad line: ' */'
./cairo-pattern.c (3346): ERROR: Will documentation comment not closed with **/
./cairo-pattern.c (3422): ERROR: _cairo_pattern_sampled_area invalid doc id (should be 'cairo_...:')
The warnings are about the documentation style used in cairo
Signed-off-by: Ravi Nanjundappa <nravi.n@samsung.com>
the active edges list must be left sorted at the next possible use
and since full_row does not deal with intersections it is not usable
when there is an intersection in the top half of the next row first
subrow
Reported-and-tested-by: Matthew Leach
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85151
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
New implementations of _cairo_pattern_sampled_area and _cairo_pattern_get_extents
which produce a more accurate bounding box. These do not depend on side-effects
of analyze_filter, can handle different horizontal and vertical scales, filters
wider than 1 for down-scaling, and compute a somewhat tighter bounding box
in most cases.
I removed the pad output of _cairo_pattern_analyze_filter as it is unused.
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
It will not use the fallback if the symbols in the previous patch
are turned on.
Also some code rearrangement to make this resemble the xlib version
more and to remove some suspect bugs. In particular meshes should not
work just because the translation is an integer.
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
Currently these are always false, but if XRender use a pixman that
supports filtering these could be turned on for that version.
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
Fallback is not used if the symbols defined in the previous patch to
indicate if XRender does GOOD/BEST are true.
This patch also includes some changes to take advantage of the fact that
if there is an integer translation analyze_filter will already have set
the filter to NEAREST.
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
Currently these are always false, but if some version of xlib uses
a pixman supporting filtering they could be changed to return true
for that version.
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
The analysis to deterimine if the GOOD filter can be replaced with
the BILINEAR filter is moved to this function so it can be used
by backends other than the image backend.
Reviewed-by: Bryce Harrington <b.harrington@samsung.com>
The pixman downscaling "95" tests attempt to rescale a 96x96 pixmap to
95x95. Ideally the borders between color areas should be sharp, but for
this use case we allow for 1 pixel of blur between the areas as
acceptable. The choice of what color to use for this blurred region is
not important, and in fact varies from backend to backend.
The old reference images were generated by Krzysztof Kosiński's
downscaling algorithm. These new images are against the algorithms
written by Bill Spitzak.
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
There are cases where the rendered output of a test can vary from
backend to backend in ways which are visually acceptable. This is why
we have reference images in the first place. In these cases, changes to
the rendering logic can result in slight differences in the output that
is also within acceptable visual limits.
We see this in the pixman downscaling tests. This script is introduced
as a way to more easily update the reference images after a renderer
change.
This script is intended to be expanded to handle updating of references
for other tests as we identify similar issues. The intent is that this
script then serves as a way to document these exceptional cases.
Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Fixes a make check error.
Checking that .libs/libcairo.so has the same symbol list as cairo.def
126a127
> cairo_lines_compare_at_y
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=84638
Signed-off-by: Bryce Harrington <b.harrington@samsung.com>
The quartz-image backend uses _cairo_surface_is_quartz(), which
therefore needs to be made available to it. Fixed as suggested by
Bryce in the referenced bugreport.
References: https://bugs.freedesktop.org/show_bug.cgi?id=84569
Signed-off-by: Andrea Canciani <ranma42@gmail.com>
When investing the symmetry of the raterisation, we want to have a
simple replay of all of the original geometry through a the flipped
recording surface. This reduces the worry about artifacts from the
clipped rendering.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As pixman uses an accumulation mask, it oversamples neighbouring edges
within a cell. We can reduce the impact of this by eliminating
overlapping triangles/trapezoids from being passed into pixman.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
