The version is 2.9 beta 2 (not in list). Best to describe with pictures: Original selection: http://i60.tinypic.com/2cpzy1g.jpg Inverted selection: http://i58.tinypic.com/kos2w.jpg Inversion seems to make a completely new selection instead of just inverting the existing one Reproducible: Always Steps to Reproduce: 1. Make a picture with some flaws around main shape and fuzzy edges. 2. Select the transparent part with the wand tool (Fuzziness = 20, rest = 0 for the above). 3. Invert selection (Ctrl+Shift+I). Actual Results: Flaws and fuzzy edges get selected. Expected Results: Flaws and fuzzy edges being outside the selection (since they belonged to the original selection). Gave it a 'major' severity, since for me personally it's an important thing to be able to get shape in the picture, _exactly_ the opposite to the one originally selected.
Hm, I suspect that some parts were selected with very low values, and since the values get inverted, they now get more selected: that's actually to be expected. A selection is a gradient from 0 to 255, and if a pixel has a selection value of '1', the inverse is 254. Would that explain what you're seeing?
(In reply to Boudewijn Rempt from comment #1) > Hm, I suspect that some parts were selected with very low values, and since > the values get inverted, they now get more selected: that's actually to be > expected. A selection is a gradient from 0 to 255, and if a pixel has a > selection value of '1', the inverse is 254. Would that explain what you're > seeing? Yeah, though still it doesn't make things any better - it's certainly correct math-wise (and helpful when increasing/decreasing selection range) but quite useless graphic-wise (no way to get the literal opposite of selection). Or perhaps it is useful, but in some means that I don't use. Lack of such trivial and basic function hurts.
Is this with the same image you used for https://bugs.kde.org/show_bug.cgi?id=343365 ?
(In reply to Boudewijn Rempt from comment #3) > Is this with the same image you used for > https://bugs.kde.org/show_bug.cgi?id=343365 ? Yup.
And was the original selection you inverted created with the magic wand tool?
(In reply to Boudewijn Rempt from comment #5) > And was the original selection you inverted created with the magic wand tool? Yup :)
Okay -- then I'm pretty sure it's the same bug, actually! Thanks for your patience -- I need Dmitry to look into the other bug, since he actually caused the regression! *** This bug has been marked as a duplicate of bug 343365 ***