[geos-devel] Assertion Failure with VC++ and STRtree:yComparator

[Charlie Savage]

Following up on my email from a couple weeks ago.

Any opinions?  What is now checked in is broken in debug builds for VC++ 
which I think is unacceptable (makes it a lot hard to run GEOS under the 
VC++ or WinDebug debuggers).

I want to revert to strk's original version ... but his comments worry me.

Charlie

Charlie Savage wrote:
> There appears to be a fairly serious issue with the trunk version of 
> GEOS with STRtree:yComparator(Boundable *a, Boundable *b).
> 
> There are various comments from mloskot and strk about how the 
> comparison should work.  Choices seem to be:
> 
>     return STRtree::centreY(aEnv) < STRtree::centreY(bEnv);
> 
> Or
> 
> return std::fabs( STRtree::centreY(aEnv) - STRtree::centreY(bEnv) ) < 1e-30
> 
> 
> Currently, the second option is used.  But on Windows, using a debug 
> VC++ build, that is almost guaranteed to cause an assertion failure. 
> VC++ has code in xutility that checks the comparison in both directions:
> 
> template<class _Pr, class _Ty1, class _Ty2> inline
>     bool __CLRCALL_OR_CDECL _Debug_lt_pred(_Pr _Pred, _Ty1& _Left, _Ty2& 
> _Right,
>         const wchar_t *_Where, unsigned int _Line)
>     {    // test if _Pred(_Left, _Right) and _Pred is strict weak ordering
>     if (!_Pred(_Left, _Right))
>         return (false);
>     else if (_Pred(_Right, _Left))
>         _DEBUG_ERROR2("invalid operator<", _Where, _Line);
>     return (true);
>     }
> 
> Assume the centers of the two envelopes are equivalent - that will 
> trigger the 2nd part of the if statement causing an assertion failure.
> 
> Can this be fixed so that the comparison works as expected?  The first 
> version does the trick, but strk wrote some scary comments that it 
> corrupts memory in some cases (although with a quick glance I don't see 
> how).
> 
> Thanks,
> 
> Charlie
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> geos-devel mailing list
> geos-devel at geos.refractions.net
> http://geos.refractions.net/mailman/listinfo/geos-devel
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3237 bytes
Desc: S/MIME Cryptographic Signature
Url : http://lists.osgeo.org/pipermail/geos-devel/attachments/20070912/07e397a5/smime.bin