[GRASS-dev] [GRASS GIS] #2583: v.net: crash on Windows

[GRASS GIS]

#2583: v.net: crash on Windows
--------------------------+-------------------------------------------------
  Reporter:  mlennert     |       Owner:  grass-dev@…              
      Type:  defect       |      Status:  closed                   
  Priority:  critical     |   Milestone:  7.0.0                    
 Component:  LibVector    |     Version:  svn-releasebranch70      
Resolution:  fixed        |    Keywords:  v.net                    
  Platform:  MSWindows 8  |         Cpu:  Unspecified              
--------------------------+-------------------------------------------------

Comment(by mmetz):

 Replying to [comment:33 annakrat]:
 > Replying to [comment:32 mmetz]:
 > > Replying to [comment:30 mlennert]:
 > > > Replying to [comment:29 wenzeslaus]:
 > > > > The test for v.net is now at 50%. This is a little bit suspicious.
 The values changed from yesterday before the last fixes (r64644, r64656).
 Test suite for v.net was added in r64639 and the values were obtained
 before the last fixes which can mean that r64644 or r64656 changed the
 results. Which results are the right ones? Is this even expected?
 > > >
 > > > MarkusM is the only one who can tell for sure, but I would not be
 surprised that changes in topology handling induce changes in primitive
 counts.
 > >
 > > The difference is caused by the fix for the uninitialized values in
 `v.net`, therefore I am sure that the new result is correct. Uninitialized
 double values (here: z coordinates) can have weird effects, I noticed that
 earlier somewhere else. Unfortunately, this bug has been present in
 `v.net` since GRASS 6.4.0 or earlier.
 >
 > OK, I updated the test with the new values.

 You can convert the old result to the new result by removing all lines
 with zero length `v.clean type=line tool=rmline`. These zero length lines
 have been caused by the uninitialized z value and are ignored in
 subsequent network analyses. That means the new result is cleaner, but the
 noise in the old result should not cause errors.

-- 
Ticket URL: <http://trac.osgeo.org/grass/ticket/2583#comment:34>
GRASS GIS <http://grass.osgeo.org>