Defect in sampling method

[jamespmonaco]

This was seen examining a svs files with jpeg compression.

If I read a 1x100 window from two neighboring x-positions in the level=1 image (i.e. downsample by 4), the results should be shifted versions of one another. They are not. Several pixel values are different.

Specifically, dest1 and dest2 should be shifted versions of one another, but are not:
openslide_read_region(osr, dest1, 0, 308, 1, 100,1)
openslide_read_region(osr, dest2, 0, 312, 1, 100,1)
dest1[1-99] does not equal dest[0-98]

This is not true for all x-positions. The problem starts for me at 312. For example, if instead we use 304 and 308, the problem does not appear.

Disclosure: I used Matlab hooks into openslide. (http://www.mathworks.com/matlabcentral/fileexchange/48887-fordanic-openslide-matlab)

[jamespmonaco]

I also confirmed this using CMU-1.svs, though problem seems to start earlier: xPos = 260