This is a combination of 2 related issues. The first and most important of the two is that the Magic Wand with 0% tolerance does not treat pixels with an alpha value of 0 the same if their RGB value is different. For instance, fully transparent black will not be selected if a pixel that is fully transparent white is selected with the 0% magic wand in add or replace mode.
The reason I noticed this functionality is due to the second issue, which is that by default when deleting a selection of pixels (be it the whole canvas or just a portion) or creating a new layer, the selection is filled with fully transparent white. However, when performing other operations such as resize image (larger or smaller), as well as using other plugins, these transparent pixels are changed to transparent black. This also commonly happens when I open a png downloaded from the internet and find the fully transparent portion of it is black, which causes issues when using the magic wand as described above, as the "default" is transparent white in paint.NET.
To give an example:
Open a new image, which should be filled with solid white by default.
Select all and delete, which fills the canvas with transparent white (the way I checked this was with the color picker tool).
Create a square of opaque white (or some other colour) within the layer.
Resize image (either smaller or larger)
Use color picker on transparent portion of image around the square, and the portion will have changed to transparent black.
Select the whole square (using rectangle select or 0% magic wand), and then move the selected pixels away from the square's current location.
Deselect the square, and then use magic wand (with 0% tolerance and add or replace) on anywhere in transparent area except from where the square used to be.
This will then highlight the void of transparent white that was left behind when the square was moved, which is not selected due to magic wand being used on the transparent black section around the old position of the square.
In addition, step 4 can be replaced with using plugins such as the latest versions of TR's AlphaCutter (which leaves transparent black behind in the areas not inside in the mask), or Vandermotten's series of Object Align functions, which leave behind transparent black in an area that represents how the object was shifted (for example if you use align to bottom and the object moves down, a thin strip of transparent black will appear at the top of the image that is proportional in size to the number of pixels the object was moved down by).
This makes using the magic wand slower than I would have hoped, because I need to use it in add mode to select the transparent black pixels as well as the default transparent white ones, in order to select all fully transparent pixels (I assume the Magic Wand would also segregate transparent red for example in this manner, which would require even more selections to be made). This needs to be done every time I do one of these operations, and the transparent black pixels then need to be deleted in order to change them to transparent white (the same is required when opening random pngs which have transparent black backgrounds).
One of the "solutions" is to increase the tolerance to around 5% in global mode, or 16% in contiguous mode on the Magic Wand to select all the transparent pixels, however I don't like doing this because it could select the very edges of the actual content on the canvas if that content has faded/feathered edges from opaque down to transparent going outwards, hence I use 0% to make sure I only select fully transparent pixels.
Fixing the magic wand issue would mask the second issue, so I don't know how relevant you feel it would be to fix the second issue also?
Are these two issues bugs, or something that I'm not quite understanding about how paint.NET works?
Thanks in advance.
