[OpenLayers-Dev] Re: [OpenLayers] #2964: Error in Layer.Zoomify when width or height is evenly divisible by 256

[alan at objectivepathology.com]

I was the submitter of the original bug. Fix works for me. I
tried zoomifying an image 16384 x 31 pixels using the
current free Zoomifyer 3.0 from zoomify.com. Note:
coordinates for zoomify tiles are z-x-y.
The result was that all the 2-n-0 tiles were 1 pixel high as
expected. All levels above that would be 0 pixels high
theoretically. However the 1-0-0 tile was 256 x 256, 0-1-0
was 256 x 256. Tiles 0-0-0 and 1-1-0 were completely
missing.
I cannot believe that such an inconsistent result is by
design.

This unfortunate result is at 
http://www.objectivepathology.com/ViewerBug/onepixelhigh.htm

In my work with pathology slides I never have a situation
like this. My only issue is the occasional scan where an
edge is exactly 256 pixels wide or high which causes a real
problem.

In conclusion, I'm happy with whatever method people choose
to handle Zoomify tiles where one dimension is zero.

Alan Page


----- Original Message Follows -----
From: "OpenLayers" <trac-20090302 at openlayers.org>
To: 
Cc: openlayers-trac at lists.osgeo.org
Subject: Re: [OpenLayers] #2964: Error in Layer.Zoomify when
width or height is evenly divisible by 256
Date: Fri, 29 Apr 2011 13:32:37 -0000

>#2964: Error in Layer.Zoomify when width or height is
>evenly divisible by 256
>---------------------+-------------------------------------
>-----------------
> Reporter:  alpage   |       Owner:  euzuro      
>     Type:  bug      |      Status:  new         
> Priority:  major    |   Milestone:  2.11 Release
>Component:  Layer    |     Version:  2.10        
> Keywords:  zoomify  |       State:              
>---------------------+-------------------------------------
>----------------- Changes (by rboulton):
>
> * cc: rboulton (added)
>
>
>Comment:
>
> I've just come across this bug myself, too.  Attached is a
>patch I came up
> with (before seeing this ticket) to fix it.  It takes a
>slightly different
> approach to the above - rather than checking for a zero
>value, it
> subtracts 1 before the % operation, then adds it again;
>this is almost
> equivalent, but if the actual tierImageSize value was 0 it
>would leave it
> at zero.
>
>-- 
>Ticket URL:
><http://trac.openlayers.org/ticket/2964#comment:1>
>OpenLayers <http://openlayers.org/> A free AJAX map viewer