[cairo] PDF backend starts to get interesting

[Carl Worth]

On Fri, 21 Apr 2006 03:28:03 -0700, Carl Worth wrote:
>
> So I've now done the work to emit this from cairo's PDF backend. The
> patchset still needs a tiny bit of cleanup before getting pushed to
> cairo's central tree,

That cleanup is done and the results are now pushed out to the central
tree.

There are currently two failures triggered by the PDF backend, though
I think neither are PDF-specific:

1) composite-integer-translate-over

	In this test, there's a half-integer translation setup,
	followed by a source surface pattern that also has a
	half-integer translation component. The net effect should be
	a transformation of the source image by an integer number of
	device pixels. But the PDF output is differing within the
	initial row and column of the output. I don't know what's
	wrong there.

2) paint-source-alpha

	This is a new test I just added, (along with caps-joins-alpha)
	for better testing of native handling of translucence.

	As far as PDF and SMask goes, this test is working
	fine. However, the result fails when poppler uses cairo to
	scale up a 4x4 image to 16x16. It appears like an image sample
	offset bug, (perhaps like the one for which there's an
	outstanding patch in bugzilla).

	This appears to be the same bug that has tripped up mozilla:

	https://bugzilla.mozilla.org/show_bug.cgi?id=324698

	I'm quite sure the librsvg guys have hit the same thing,
	(though I'm not currently finding the right bugzilla entry for
	that).

	So, even though it's not related to what we're trying to test
	here, I'm leaving this test as a failure to help remind us to
	fix it.

Meanwhile, I took about 5 tests that were using ARGB32 source surface
patterns with the alpha channel set to all 1.0 and changed them to use
RGB24 surfaces instead. This didn't change the semantics of the tests
at all, but did make the PS and PDF backends start emitting native
rather than fallback output for these tests.

I did this to exercise a bug that I was fixing in the PDF backend,
(with regard to the transformation of source surfaces). But it ended
up uncovering a bug in the PS backend as well. So there are currently
5 test failures in the PS backend that need to be fixed, (likely just
one bug there).

-Carl
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.freedesktop.org/archives/cairo/attachments/20060425/5e07a332/attachment.pgp