
Hey Devs,
I've just pushed r12525 which should fix an error where moving filtered items wasn't using the right bbox.
http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/revision/12525
What concerns me is two things: That this was effecting text elements, but not other items suggests that _bbox and _item_bbox were different for text but not for other items. Which is odd and should probably be fixed too.
The other is that I've removed the now unused compute_drawbox. It mentioned converting to user units and I may have inadvertently hurt something there by using area_enlarge instead. Any unit people want to confirm this is safe or unsafe?
And I didn't even get to fix the original bug I was after. ;-)
Best Regards, Martin Owens

2013/9/17 Martin Owens <doctormo@...400...>:
The other is that I've removed the now unused compute_drawbox. It mentioned converting to user units and I may have inadvertently hurt something there by using area_enlarge instead. Any unit people want to confirm this is safe or unsafe?
You can test this by creating a rotated group and putting a filtered object inside it. You might want to do this manually in the XML editor, because AFAIK Inkscape will push the transforms down to the lowest possible level (so e.g. rotating a group with rectangles will cause the transform attributes to be set on the rectangles, not on the group).
The filter rendering code contains serious design errors, as it does not use the same area invalidation model as the rest of the drawing tree. As a result, some things - such as implementing support for filter primitive areas - are not possible. I didn't manage to fix it during the Cairo conversion, because I had a really hard time wrapping my head around all the different coordinate systems being used.
Regards, Krzysztof
participants (3)
-
Diederik van Lierop
-
Krzysztof Kosiński
-
Martin Owens