trunk: poor rendering of path effects after 2geom update (rev >= 12921)

[su_v]

Steps to reproduce:
1) launch current trunk
2) open 'share/examples/live-path-effects-gears.svg'

--> the quality of the computed path of the live path effect with r12921 has deteriorated compared to r12920 and current stable. The regression affects on-canvas rendering as well as bitmap export, and persists if converting the path effect to a static path (Shift+Ctrl+C).

First encountered while testing r12975 on OS X 10.7.5.

Based on tests with archived builds (on OS X 10.7.5):
- not reproduced with rev <= 12920
- reproduced with rev >= 12921
this regression was introduced with the recent update of lib2geom in revision 12921:
<http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/revision/12921>

Also affects:
- Bend LPE
- Envelope Deformation LPE
- Ruler LPE
- possibly other LPE (more tests needed)

[su_v (suv-lp)]

Bitmap export (drawing area) with r12920

[su_v (suv-lp)]

Bitmap export (drawing area) with r12921

[su_v (suv-lp)]

Regressions with ruler lpe after the 2geom update (rev >= 12921).

[su_v (suv-lp)]

Regression with bend lpe (rev >= 12921)

[su_v (suv-lp)]

Variation of bend lpe regression (no transforms on parent container, all orig paths with the same precision).

[Alvin Penner (apenner)]

confirmed on Windows XP, Inkscape rev 12954

I have looked only at the gear path effect, but it appears that the defects are associated with Bezier control arms that are collinear. If I convert to static path and click on endpoints, it seems that the defective segments have Bezier control arms are always parallel (0 degrees) or anti-parallel (180 degrees) and sometimes reversing the direction of the control arm improves things.

[Alvin Penner (apenner)]

probably related to a change I made in lib2geom rev 2122.
I'll look into it.

[Johan Engelen (johanengelen)]

Thank you very much for looking into this Alvin

[Alvin Penner (apenner)]

fix committed to rev 12980, added support for case where a Bezier control arm length is zero.

normally this should be committed to 2geom first, but I thought I would commit it here for testing purposes, and if it survives testing for a few days, then I'll commit it to 2geom as well.

[Johan Engelen (johanengelen)]

great Alvin.
Don't take the 2geom-commit-rule too strict :-) Just don't forget to do it !!!

[Alvin Penner (apenner)]

committed to lib2geom rev 2164