When running the tests, only run the random (and potentially large) size
download test once instead of 10 times.
There's no real benefit in doing that, both because it's unlikely to
fail only in the 2nd or 9th run and because the sizes are picked
randomly.
This also speeds up the test massively as the download test was
dominating the runtime.
Instead of picking a few numbers in advance and running them through the
test gauntlet every time, pick the random numbers at runtime.
This both increases the test coverage in that it ultimately tests more
combinations across many runs and it reduces the runtime of individual
runs because every tun only runs the download tests twice (with 1px and
the random size) instead of 5 times.
And that speedup benefits the CI, where the asan runs would cause this
test to timeout sometimes.
Add some odd-sized texture sizes to the
download tests, to trigger alignment issues
in the various upload code paths. And add
a size that is bigger than the max-texture-size
we force in one of our test setups.
To compensate, reduce the number of
runs per size from 20 to 10.
The GL renderers like to premultiply content that isn't, and due to the
data loss with alpha == 0 (transparent white, transparent black and
transparent anything are all represented by (0, 0, 0, 0) when
premultiplied) these values cannot be converted back.
There is no longer a need to use gdk_texture_download() and force
conversion to ARGB8 format. We can download the pixels in the original
format again.
That way we avoid testing the conversion code and avoid having to deal
with differences in representable colors.
However, some formats do do conversions, so we allow pixel comparisons
to be accurate (requires 16bit comparison accuracy) or inaccurate (we
only care about 8bit).
Note that for the default RGBA formats, this is identical and means they
need to be bit-exact the same, no matter what.
But the higher bit depth formats may be more different - floating point
can even have different values with high accuracy (the float mantissa is
23 bit, we only care about 16).
Add some tests for handling of failures.
The test data here is taking from gdk-pixbufs
tests/test-images/fail directory, excluding anything
but png, tiff and jpg images.
These tests can be run manually, but are not suitable for use as an
acceptance test, so let's not make frameworks like Debian's autopkgtest
run these when they run ginsttest-runner in the most obvious way.
Signed-off-by: Simon McVittie <smcv@debian.org>
There are two possible interpretations of "expected failure": either
the test *must* fail (exactly the inverse of an ordinary test, with
success becoming failure and failure becoming success), or the test
*may* fail (with success intended, but failure possible in some
environments). Autotools had the second interpretation, which seems
more useful in practice, but Meson has the first.
Instead of using should_fail, we can put the tests in one of two new
suites: "flaky" is intended for tests that succeed or fail unpredictably
according to the test environment or chance, while "failing" is for
tests that ought to succeed but currently never do as a result of a
bug or missing functionality. With a sufficiently new version of Meson,
the flaky and failing tests are not run by default, but can be requested
by running a setup that does not exclude them, with a command like:
meson test --setup=x11_unstable --suite=flaky --suite=failing
As a bonus, now that we're setting up setups and their excluded suites
programmatically, the gsk-compare-broadway tests are also excluded by
default when running the test setup for a non-broadway backend.
When running the tests in CI, --suite=gtk overrides the default
exclude_suites, so we have to specify --no-suite=flaky and
--no-suite=failing explicitly.
This arrangement is inspired by GNOME/glib!2987, which was contributed
by Marco Trevisan.
Signed-off-by: Simon McVittie <smcv@debian.org>
That way we ensure that the GL context(s) get disposed, which they
previously weren't due to them still being the current context.
This also implicitly adds testing of gLContext destruction, which
previously wasn't ever done by any test.
