[geos-devel] test failures

[Sandro Santilli]

On Sat, Aug 17, 2013 at 09:06:34PM -0400, Greg Troxel wrote:
> 
> With 3.4.1, I ran the tests on NetBSD 6, i386.  I got 3 failures. Are
> these really "XFAIL" as some test driveres woudl call them, and known
> bugs and not indicative that anything is worse on my platform than
> anywhere else?

I'm afraid yes. They seem to give slightly different results than
expected, but the bug they refer to was to not return a result at all
or (in the case of the bogus noding) not returning a validly noded
result.

So, actually, only the last one should be reviewed to check if the
output is correctly noded, but I think node checking is one of the
things done better in 3.4.0.

The real fix for this would be to implement pluggable test checker
in the XML test runner (like it is done in JTS) and use a better
test for comparing expected vs. obtained result. Specifically a test
for the noding in the case of the noding bug and a test for similarity
in the others (we do have HausdorffDistance test that would help in there).

Any volunteers to help there ?

--strk;