On Sat, 18 Jan 2025 11:49:55 +0000 Simon McVittie <s...@debian.org> wrote:
On Sat, 18 Jan 2025 at 11:39:41 +0000, Simon McVittie wrote:
> > 65/693 gtk:gsk+gsk-nodeparser / parser empty-texture.node
FAIL 0.03s exit status 1
> > 66/693 gtk:gsk+gsk-nodeparser / parser empty-texture-scale.node
FAIL 0.03s exit status 1
> > 94/693 gtk:gsk+gsk-nodeparser / parser texture-fail.node
FAIL 0.03s exit status 1
> > 95/693 gtk:gsk+gsk-nodeparser / parser texture-names.node
FAIL 0.02s exit status 1
> > 96/693 gtk:gsk+gsk-nodeparser / parser texture-scale-filters.node
FAIL 0.03s exit status 1
> > 97/693 gtk:gsk+gsk-nodeparser / parser texture-scale-unknown-filter.node
FAIL 0.03s exit status 1
>
> These are "reftests" which draw the same scene in two different ways,
> and assert that the node tree in GSK's scene graph is equivalent. With
> libpng 1.6.44, they pass.
Sorry, pressed send too soon. What I meant to say for the rest of that
paragraph is:
If I'm understanding correctly, with libpng 1.6.45, the small texture
that is involved in these tests (which is serialized as a data: URI
containing a base64-encoded PNG) is different in the serializations of
the two scene graphs - probably a colour space difference, the same as
the other failing test? - and as a result the test fails.
smcv
Hello, looks like this change was intended.
gtk people looks already fixing the test, maybe we can ignore them for now?
G.
