gradiant error on mitred joins

[Richard]

I tried to draw a rectangular overlay consisting of four trapeziums arranged to fit together with mitred (or mitered in US) 45 degree joints (arranged like the wood in a picture frame in other words). If this messaging system preserves the line spacing then the shape (in outline) looks something like this below...
_____
\____/

The trapezoids have no border. Each one contained a gradiant fill flowing from opaque black at the outside (longer edge) to transparent black at the inside (short edge). At the joins one would expect the gradiants to track each other - dark at the outside and see-through at the inside. They don't. Firstly there is a thin transparent separation between the shapes. In addition the fill gets slightly more transparent toward the mitred edges meaning that while the gradiant function defines the graduation as occuring in one plane it actually varies in two planes - and this is the bug I believe. The effect is that when this is exported as a .png bitmap there is a diagonal line of less opaque pixels at each mitred joint surrounded by a region of enhanced transparency (however transparent the image is meant to be then as you get closer to the corner it is more transparent). While it sounds trivial the effect is extremely noticeable.

The shapes were approximately exactly 2000 pixels by 1000pixels when constructed into their frame-type arrangement.

I hope that makes some kind of sense.

Thanks,

Richard

[Liam P. White (liampwhite)]

Correct me if I'm wrong, but as far as I can tell, this behavior has been reported many times before. See screenshot.

If this is what you are describing, then I propose to link as a duplicate of bug #170356 "
Provide per-object antialiasing control"

[Richard (rs-q)]

Yes Liam I think it probably is the same issue that is being discussed. I am happy that you combine this. I had a look for similar issues when I originally reported but the symptoms described here were a long way from my description which is why my search did not succeed. That's life I guess.

I hope that you don't like me saying that I don't particularly like the title of 170356 because it is suggesting a cure and I would not want to do that when reporting a bug. At that point I just want to let people know that there is an issue and not how to fix it. Why? Because that issue might have many proposed cures and I would be no judge as to which are ones valid and of those then which cure would be best. Better people than me do that. Does that make sense?

Thank you.

[su_v (suv-lp)]

Linking as duplicate to bug #170356 as discussed (same underlying issue).