MysteriousBugOracle's profile - activity

Using an experimental new static analysis technique we (GrammaTech) detected what may be a bug in createIndicesDists, miniflann.cpp:

static void createIndicesDists(OutputArray _indices, OutputArray _dists, 
                                       Mat& indices, Mat& dists, int rows, 
                                       int minCols, int maxCols, int dtype) 
        { 
            if( _indices.needed() ) 
            { 
                indices = _indices.getMat(); 
                if( !indices.isContinuous() || indices.type() != CV_32S || 
                    indices.rows != rows || indices.cols < minCols || indices.cols > maxCols ) 
                { 
                    if( !indices.isContinuous() ) 
                       _indices.release(); 
                    _indices.create( rows, minCols, CV_32S ); 
                    indices = _indices.getMat(); 
                } 
            } 
            else 
                indices.create( rows, minCols, CV_32S ); 

            if( _dists.needed() ) 
            { 
                dists = _dists.getMat(); 
                if( !dists.isContinuous() || dists.type() != dtype || 
                   dists.rows != rows || dists.cols < minCols || dists.cols > maxCols ) 
                { 
                    if( !indices.isContinuous() ) // HERE!  should this be 'dists'?
                        _dists.release(); 
                    _dists.create( rows, minCols, dtype ); 
                    dists = _dists.getMat(); 
                } 
            } 
            else 
                dists.create( rows, minCols, dtype ); 
        }

The block of code operating on _dists/dists appears to be a copy of the above block operating on _indices/indices, with all of the occurrences of _indices/indices changed to _dists/dists except the one noted in the "HERE!" comment. This inconsistency couid be intentional, but it is highly suspicious. Can anyone "in the know" confirm or deny that this is a bug?

Factoring? // a couple of suitable typedefs... static void cvImageWidget_get_preferred_dimension (Gtk

This could have been avoided by a little factoring, as follows: // there might be a better name for this... static void

Do I need to create another separate account for your bug tracking system, or does the one used for this site cover both

We don't actually use this code, so we're not in a position to create or test a fix. We just run our static analysis to

Possible bug in modules/calib3d/test/test_stereomatching.cpp? Using an experimental new analysis technique, we (GrammaTe

Possible bug in modules/highgui/src/window_gtk.cpp Using an experimental new static analysis technique, we (GrammaTech)

is this code in modules/flann/src/miniflann.cpp a bug? Using an experimental new static analysis technique we (GrammaTec