Subject: [ggl] Difference / xor for multi polygons
From: Barend Gehrels (barend)
Date: 2011-03-05 09:13:22
> Randomized testing of manhattan data validated against Polygon's result would have caught this problem. The fact that the bug is now fixed does not ensure that similar problems do not still exist. The testing I'm suggesting is a form of pseudo-exhaustive verification. All possible cases can't, of course, be tested, but with randomized testing you can quickly and easily test a huge number of cases, enough to conclude that there are no latent bugs. From my point of view, performing this kind of verification before release of the library is the responsible thing to do. Any bugs found after the release will reflect upon the entire boost community. The worst thing that can happen is that after a few hours work you find no issues and can congratulate yourself on a job well done. Wouldn't you rather know for sure that you code is correct?
As said, we have many random tests AND structural tests. Testing huge
number of cases.
Anyway, some verification is always useful, so I started. But it turns
out to be useless, in an early phase Boost.Polygon crashes or gives me
the wrong results. I've reported them to the Boost mailing list as it
does not belong to GGL. Please answer there and not here as this is the
GGL mailing list.
So thanks again, but I stay to automatic self-verifications; or test
with SQL Server Spatial, which literally always gives the right answer.
Geometry list run by mateusz at loskot.net