One test has all properties set to 'inherit', the other to 'initial'.
This should result in the same result, as inherit will ensure every
widget inherits the same value, and for the toplevel 'inherit' is
defined as being identical to 'initial'.
From the css docs at http://www.w3.org/TR/CSS2/box.html:
8.5.1 Border width: 'border-top-width', 'border-right-width', 'border-bottom-width',
'border-left-width', and 'border-width'
Computed value: absolute length; '0' if the border style is 'none' or 'hidden'
So, if i specify border-style none and a border-width > 0 that should give the
same result as border-width 0.
We get rendering artifacts that make tests fail. Not good.
The code is only commented out so far, so it should be trivial to
reenable if someone wants to. (Or this commit could just be reverted.)
row-spacing behaved like column-spacing should have and vice versa.
Also update tests that erroneously checked this behavior.
Thanks to Joanmarie Diggs for finding this.
When tests are larger than the screen size and no compositing is
enabled, the window will be clipped to the screen size and all areas
outside of the screen have undefined contents.
To avoid this, we can use an offscreen window.
This test is copied from the CSS spec. The reference is very sensitive
to the rendering algorithm in use, because the rounded corners have
semi-transparent pixels. It might be necessary to update them in the
future.
Tests that border-image-repeat is rendered correctly.
The reftest is done via toolbars stuffed into a GtkFixed using
background-image and images tuned for it.
We need pay attention the the request mode when doing size allocation.
The code was using request mode for requisition, but orientation
for allocation.
Also add a reftest that exhibits the issue, courtesy of
Benjamin Otte.
In widget hierarchy like "Foo Bar Bar Baz", we want the selector "Foo >
Bar Baz" to match, because it matches the elements 1, 2 and 4.
Previously, the selector only matches the Bar at position 3 and then
failed because it wasn't preceded by a Foo.
See the documentation in the script.
Tests are not yet added as the output from the 2 included scripts
doesn't match and the intended reference output first needs to be agreed
on.
GtkArrow and the align properties use different methods (float vs int)
to center the arrow. If the size of the arrow is odd, this will cause a
rendering that differs by half a pixel. So we request an even size for
both the arrow and the container and everything works out.
If we have a toplevel, and not a popup window, do wait an additional
0.5s to give the WM/server enough time to actually create the window.
This is a hack and there should be a better solution. But it works.
Please use POPUP windows for tests unless the test must use toplevel
windows.
Checks that the size requests for labels are as they should be for
required and natural size given various combinations of wrap, ellipsize,
width-chars and max-width-chars.
See
http://mail.gnome.org/archives/gtk-devel-list/2011-April/msg00036.html
for the discussion spawning this test.
Add a new test runner supposed to do a lot of generic tests. Run it like
this:
./gtk-reftest [OPTIONS] TESTFILE [TESTFILES...]
where FILE is a GtkBuilder ui file to run.
For a general test named "test", you want to have the following files:
1) test.ui
2) test.ref.ui
3) test.css (optional)
The test will then check that test.ui and test.ref.ui are rendered
identically with the provided css.
In detail, for every provided TESTFILE the test runner will:
1) Add the css to the default screen
2) Load the test.ui file and the test.ref.ui file
3) Grab the first GtkWindow subclass widget
4) gtk_widget_show() it and take a snapshot image of its contents into
a cairo surface.
5) Compare the two images to be bitwise identical. If they are not, a
diff image will be created hilighting the differences.
6) Save the images as png files to the output directory named:
- test.out.png (rendering of test.ui)
- test.ref.png (rendering of test.ref.ui)
- test.diff.png (optional, differences from step 5)
7) Fail the test if the two images are not bitwise identical
Credit for the idea of reftests goes to Mozilla and in particular David
Baron. For a larger introduction of why reftests are useful, see
http://weblogs.mozillazine.org/roc/archives/2008/12/reftests.html
