On Fri, Sep 20, 2013 at 06:25:27PM +0200, Krzysztof Kosiński wrote:

2013/9/20 Bryce Harrington <bryce@...961...>:

...

On Tue, Sep 10, 2013 at 11:08:34AM +0200, Tavmjong Bah wrote:

...

...

On Mon, Sep 9, 2013 at 4:32 PM, Martin Owens <doctormo@...400...> wrote:

...

On Mon, 2013-09-09 at 15:46 -0700, Josh Andler wrote:

...

In addition too that, we will need to bring a copy of cairo and pixman in-tree for a while to ensure the new renderer will work everywhere.

I don't really like the idea of bringing cairo and pixman into the tree. For Windows and Mac, aren't they already included somehow? For Linux, by the time of release, the distros should already be up-to-date with the latest Cairo assuming that 1.14 is released in mid-October as planned (see cairo mailing list 5 Sept).

Is it only the downscaling bug that necessitates bringing these in-tree? Or are there other bugs needing new cairo/pixman features?

Also, anyone happen to have a good demonstration of the problem they should show me? I'd like to improve the test case I wrote for this, and would help to see some real-world examples of where people have run into the bug.

This is the relevant Inkscape bug: https://bugs.launchpad.net/inkscape/+bug/804162

I don't have a Cairo-only test case, but I could make one if required. Draw some 1-pixel lines on an image surface and then paint it scaled down by a factor of at least 2 to a second surface. With Cairo 1.12, the lines will have gaps or disappear, depending on their angle and position. With Cairo 1.13, they should show correctly.

Thanks, I was able to create a test case that reproduced those bugs with 1 pixel lines. Looks like the lines disappear for specific scaling factors (e.g. 0.25 and 0.33 but not 0.4). I'm considering replacing my earlier test case with this new one since it tests the bug more precisely.

Bryce