As anyone who uses Inkscape for real work knows, stroked shapes are prone to becoming invisible at certain zooms. It's especially bad for complex path, and playing with L-system effect recently made it so annoying for me that I searched for a fix. Which turned out to be quite trivial, so I committed it. This closes at least two bugs:

http://sourceforge.net/tracker/index.php?func=detail&aid=1244737&gro... http://sourceforge.net/tracker/index.php?func=detail&aid=1254932&gro...

and a related bug where a complex path actually disappears (not just becomes invisible) after an offset:

http://sourceforge.net/tracker/index.php?func=detail&aid=1498546&gro...

Of course this change (in livarot lib) is somewhat risky, but in my testing it works well. Very complex paths that would previously simply disappear, now may display certain small rendering artefacts (horizontal lines), but i think it's much better than becoming totally invisible. Please test thoroughly; if you see something bad, especially crashes, I would back this out.

For those interested, the reason for this bug was given by Fred in this comment:

shape_euler_err = 4, // computations result in a non-eulerian graph, thus the function cannot do a proper polygon // despite the rounding sheme, this still happen with uber-complex graphs // note that coordinates are stored in double => double precision for the computation is not even // enough to get exact results (need quadruple precision, i think).

So in most cases this error was caused by rounding errors, not by the shape graph really being non-Eulerian (which means some nodes have more in-coming edges than out-coming edges). So I simply removed the check for Eulerianness for stroke shape. My understanding is that even if we hit some shape which is not Eulerian, the renderer will simply do some artefacts but should not crash.