A slight variation on the simple fill-alpha theme is that if Company is
truly seeing a regression where we dither, the error on a constant
background will be small and might be missed by the test suite.
Benjamin Otte reported "who broke rgba fills? they look dithered with
recent git."
This commit is the result of a skim through the test-suite which revealed
no single test responsible for checking the basic operation of
"set_rgba(); fill();".
If the gradient has constant alpha, then we can express it as a flattened
linear gradient. Otherwise, should the opacity vary across the gradient we
need to fallback.
Quick summary of changes:
- Move list of cairo source files out of src/Makefile.am and into
src/Sources.mk,
- Generate files src/Config.mk and src/Config.mk.win32 that choose
the right set of source files and headers based on configured
backends and features. This drastically simplifies building
using other build systems. The src/Makefile.win32 file needs
to be updated to reflect these changes.
- Add README files to various directories,
- Add toplevel HACKING file.
Be explicit about handling cached FAIL images, instead of relying on the
sequences of failed matches as the files are an external resource and we
can not guarantee their individual accessibility.
Note this also changes the filename, so you may want to run:
$ find -name '*-last.*' -print | xargs rm
after this checkout.
Compare the current output against a previous run to determine if there
has been any change since last time, and only run through imagediff if
there has been. For the vector surfaces, we can check the vector output
first and potentially skip the rasterisation. On my machine this reduces
the time for a second run from 6 minutes to 2m30s. As most of the time,
most test output will remain unchanged, so this seems to be a big win. On
unix systems, hard linking is used to reduce the amount of storage space
required - others will see about a three-fold increase in the amount of
disk used. The directory continues to be a stress test for file selectors.
In order to reduce the changes between runs, the current time is no longer
written to the PNG files (justified by that it only exists as a debugging
aid) and the boilerplate tweaks the PS surface so that the creation date
is fixed. To fully realise the benefits here, we need to strip the
creation time from all the reference images...
The biggest problem with using the caches is that different runs of the
test suite can go through different code paths, introducing potential
Heisenbergs. If you suspect that caching is interfering with the test
results, use 'make -C test clean-caches check'.
In order to achieve substantial speed improvements the external conversion
utilities are rewritten as a daemon that communicates with the test suite
over a local socket. This is faster as it avoids the libtool and dynamic
linker overhead for each invocation, the caches persist between tests and
we no longer require a round trip through libpng.
The daemon is started automatically by the test suite and if communication
cannot be established then it falls back to using a pipe to a normal
conversion utility. The daemon will then persist for 60 seconds waiting
for further connections.
Of course any memory leak (stares at poppler) is exacerbated.
Franz Schmid reported on the mailing list,
http://lists.cairographics.org/archives/cairo/2008-April/013912.html,
an issue with drawing a dashed rubber band in Scribus. The problem
appears when segments of the dashed rectangle are outside the image,
with the errant dash connecting the ends of the visible segments.
This bug first was fixed in 40558cb15e,
but then reintroduced in 9cfd82e87b, which
became part of the 1.6.2 quick release.
As penance to make sure I never repeat this same bug again, I offer this
test case which exercises the XSetClipMask(NULL) path and hopefully
simulates some 'typical' usage of cairo by GUI toolkits.
The experience of running the entire test suite under cairo_push_group()
did at least reveal that there was a potential bug in the pdf backend.
This test is just the simplest of drawing operations - simply filling
the similar surface with solid blue, and then blitting that surface to
the destination.
This test exercises code that computes the extents of a surface
pattern with CAIRO_FILTER_BILINEAR, (where the filtering
effectively increases the extents of the pattern).
The original bug was reported by Owen Taylor here:
bad clipping with EXTEND_NONE
http://bugs.freedesktop.org/show_bug.cgi?id=15349
PDF has a concept of "soft" masks, for which it is able to construct a
mask out of PDF drawing operations. These tests exercise constructing
various masks using the high level drawing operations.
Interestingly, this test case does demonstrate that cairo-pdf
is fixed, (where without commit f6509933a4 the Y positions
of the glyphs were inverted); however, cairo-ps is failing
with this test, (all the glyphs are ending up on top of each
other).
This reproduces the bug in PS/PDF/Win32-printing where an unsupported
operation that is off the page causes an unsupported operation
assertion in the backend.
Parts of the stroker depend upon whether we have in effect a reflection
matrix (one whose determinant < 0). This test incopororates the same
drawing under the a couple of reflections to exercise stroking under
matrices with both positive and negative determinants.
Draw a few rectangles whose vertices are outside the bounds of the
surface, but whose segments cross the surface. This exercises the new
dashed stroker optimisation which tries to determine whether the line
segment is visible.
Inspired by bug 7362 (painting a glitz surface onto an xlib surface
crashes cairo) and the lack of coverage for
_cairo_paginated_surface_acquire_source_image(), these tests attempt
to use each backend as a source surface for all the other backends.
For example, this checks that one can construct a PS file ready for
printing and then copy that surface to an image/xlib for previewing.
Add various test cases to exercise
_cairo_pattern_acquire_surface_for_surface(), most notably using similar
source surfaces to provide coverage of the non-image surface branch.
This demonstrates an error in cairo where miter joins are replaced with
bevels at high scale factors due to a test added to eliminate wild miters
drawn when the line faces are nearly parallel.
This test demonstrates a bug when compositing an rgb24 image over an argb32
image, (the implementation appears to be examining the alpha channel
rather than ignoring it).
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.)
