Boost C++ Libraries: Ticket #10976: Problem with polygon library versions 1.56 and 1.57

Wed, 28 Jan 2015 09:05:49 GMT

Hi Luke, tried to contact you by email but it bounced back. So I chode this way, regards, Uwe.

owner changed

Andrii Sydorchuk — Sun, 08 Feb 2015 22:12:34 GMT

owner changed from Lucanus Simonson to Andrii Sydorchuk

cc changed

Fri, 06 Mar 2015 09:34:48 GMT

cc bjornpiltz@… added

Andrii Sydorchuk — Fri, 20 Mar 2015 20:49:13 GMT

Hi Uwe,

Thank you for your report. Indeed, the regression was introduced in Boost 1.56. I applied the fix and going to add unit tests to cover the issue. The fix will be merged into release branch of Boost 1.58.

Andrii Sydorchuk — Fri, 20 Mar 2015 21:44:52 GMT

The fix with the corresponding test was pushed to the develop branch.

status changed; resolution set

Andrii Sydorchuk — Fri, 20 Mar 2015 21:55:37 GMT

status new → closed
resolution → fixed

The fix was pushed to the master branch.

mhilferink@… — Tue, 30 Jun 2015 10:50:14 GMT

Note that insert_vertex_sequence() calls insert_clean(elem) which has is_hole as a second optional parameter and already has code to reverse the orientation. It would be cleaner to pass is_hole to insert_clean and avoid scattering the orientation logic around.

BTW, why is on various places the orientation reversed if a edge is horizontal or vertical? I can see that intersected segments can have a reversed order of first and second point due to round off, making a vertical (S<-N) segment from an (NW->SE) edge, or a (S->N) segment from a (SE<-NW) edge. But reversing on specific directions such as in validate_scan, independent of whether point order did change or the points were actually swapped, seems something that is attempted to be compensated elsewhere, such as in _processEvent.