I am the upstream maintainer of resynthesizer.
I think the resynthesizer behaviour you describe is as designed. In
other words, this is not a bug. The test case is from an obsolete
tutorial, obsolete because of intended changes to resynthesizer.
My understanding of your test case:
You followed the tutorial you linked. In that tutorial, you use GIMP
menu: Filters>...>Resynthesize. That is what I call a raw interface to
the resynthesizer engine: it lets you fully control the parameters of
the engine. If you use the GIMP menu: Filters>Enhance>Heal Selection
you should get the results I think you expect. That is a plugin that
adapts to the engine by providing a simpler GUI and sending the
appropriate parameters to the engine.
What has changed:
The old version of the resynthesizer used the inverse of the selection
as the source for synthesizing the selection. The new version does not,
and more or less requires a separate parameter to specify the source for
resynthesizing (usually a separate image layer.) Thus in the obsolete
tutorial test case, you resynthesized the watermark from itself, and the
result seemed unchanged.
Discussion:
It may have been a design error to change the resynthesizer as
described. It has led to some confusion such as this bug report. The
intention of the change was to simplify the code and API. The fact that
the old engine was inverting the selection mask to derive the source of
resynthesis was not known by most users, and in my opinion, was itself a
source of confusion. The new API says: here is an image with the target
selected, and separately, here is another image with the source
selected. The old API said: here is an image with the target selected,
and in the absence of a separate parameter for the source, the engine
will use as source the inverse of the selection in the same image.
I don't think it would help to revert the engine to the old API. I hope
that eventually the confusion will cease.
Note that the signature of the API has not changed, only the behaviour.
I could be wrong, if you still suspect a flaw, please don't hesitate to
report it.
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
