On Thu, Jul 21, 2005 at 08:16:27PM -0300, bulia byak wrote:
On 7/21/05, Kees Cook <inkscape@...62...> wrote:
[pjrm] Possible W3C test regressions.
Status: Peter needs to take a look and fix if easy. https://sourceforge.net/tracker/?func=detail&atid=604306&aid=1224751... Bryce and Kees looked at this: doesn't entirely look like a show-stopper.
I have just tested and I don't see any regressions. Rick Beton: are you sure you tested the latest build? Can you try again?
All styling-* tests work. I also tested a few others in text and paths, and found some more that are pass though you marked them as fail or partial. I updated the wiki page. We are now at 69/25/86 pass/partial/fail.
Carl showed me how he's been running the W3C suite for his cairo development. Essentially he renders all of the test cases to png, and then generates a visual diff and reviews it to see how well it does. the comparisons are done against his last "accepted" stable version. This way he can track as the compliance improves.
I got a copy of Gaze from Josh for Carl to use as a performance test on Cairo. Gaze is 10mb and pretty intensive to render. I was able to render it pretty well on my laptop, although zoming around took a noticeable time to render. On Cairo, however, it took quite some time to render (a few minutes I think), and the rendering was quite different than it was in inkscape. It's not clear if the differences are due to cairo or inkscape, but I think this will help work towards convergence.
In addition to the W3C test suite, I think we can also gain value from fishing out all the failed clipart from the OCAL tests, and start building a regression test set from them. Carl's also interested in this, and I bet others will as well.
Btw, librsvg has added an ability to switch out backends, and is now able to render using Cairo. Some of the guys here were playing with doing Cairo rendering for Gtk widgets.
Bryce