I've been hacking at the internals of the transform tool, but until now I hadn't looked at how it worked from the outside. It's just not right! You need to be able to set the transform by swapping between modes (rotate/shear/scale/perspective - plus move), adding to the existing transform at each stage, and checking the result (probably with no smoothing, for speed). Once it's right, the final transform should be done in one step from the original layer. Right now, it's not very useful for real work. (Can some users back me on this?) This is independent of how it handles selections, of course. (As an aside, I think 1.0.4 was closer to the correct behaviour, at least in performing each transform from scratch instead of from the previous result. Do we need some kind of forum for discussing and keeping track of design decisions like these? Preferably one with working artists involved. Or are they listening here?) So, does this fall inside or outside the feature freeze? (It's not exactly new code, mostly just rearrangement.) Is anyone else in a position to work on this, or should I take it on? Does anyone seriously disagree that this needs doing, even if not for 1.2? -- David Hodson -- hodsond@xxxxxxx -- this night wounds time
