1 pixel shift in subsequent rasimports

[beccati@…]

There is a shift of one pixel along the y axis for subsequent imports of the same image repeated along the z axis. The first z layer is shifted with respect to the subsequent ones (or the other way around).
This shift appears only when more than one image is loaded into the same layer.

The attached test case loads repeatedly a single image then computes md5 sums of a given section extracted in several layer. Then it loads image pairs and the md5 difference can be seen.
There is also a screenshot attached showing the image differences with a point at same coordinates (first insert on the left)

Alan

[beccati@…]

test case for shifted inserts

[herziga@…]

The behaviour Alan described is a bug and is caused by a rounding issue in conjunction with negative shift vectors in x- or y- direction. I've uploaded a patch to fix the issue.

Alex

[beccati@…]

Where can I find the patch? Or has it already been committed to the git repository?

Alan

[herziga@…]

Replying to beccati@…:

The patch hasn't been applied to the core but you can download it anyway from the patchmanager section on the website: ​http://rasdaman.eecs.jacobs-university.de/trac/rasdaman/patchmanager
Just click on the green arrow in the table.

Alex

Where can I find the patch? Or has it already been committed to the git repository?

Alan

[beccati@…]

The patch solves the issue, now the md5 are all the same and they are different from both the previous ones indicating a consistent different shift

Alan