[geos-devel] Current Status

strk strk at keybit.net
Wed Apr 28 09:59:46 EDT 2004

On Fri, Apr 23, 2004 at 09:56:31PM -0400, Carl Anderson wrote:
> Is the current API creep in CVS intended?
> toGeometry(Envelope *, ... ) has changed and no long compiles against  
> PostGIS cvs nor PostGIS 0.8.1
> I ask to see where I should try to plug bugs?
> the CVS code is more delicate (has more instance of SIGABRTs) than the  
> geos-1.0 code.
> The CVS work has promising elements in it but while the 1.0 code
> only SIGABRT'ed on 15 of 200,000 geometries the CVS code
> SIGABRTs on more than 200 of same set.
> The 1.0 code only
>    SIGABRTs in the BufferOp routine
> and CVS
>    SIGABRTs in noding/IteratedNoder.cpp and strtree and other places  
> but not BufferOp

I've found the IteratedNoder SIGABRT cause:
  AssertionFailedException: Should never reach here: AbstractSTRtree::query encountered an unsupported childBoundable type

Since IteratedNoder::node is a recursive function, the exception mechanism
might not be able to realize who should get the thrown exception.
I've tryed catching the exception in IteratedNoder::node itself, printing
it to the screen and re-throw it again.
Then from BufferBuilder::computeNodedEdges I've tryed to catch that
same exception (it calls IteratedNoder::Node) but I get a SIGABORT
before being able to print the exception.

Beside fixing the exception cause, does anyone have an hint on
how to handle this ? 
Can you confirm this behaviour ?
Should we use a special return value ?


> remembe that my test set is by definition not unlimited and other  
> instances of trouble may be waiting in both CVS and 1.0.
> so should I chill and wait for CVs to settle, work on submitting a  
> patch against 1.0 or work on CVS.
> If the CVS API changes are intended could someone increment the version  
> tag (and directory name).
> C.
> _______________________________________________
> geos-devel mailing list
> geos-devel at geos.refractions.net
> http://geos.refractions.net/mailman/listinfo/geos-devel

More information about the geos-devel mailing list