CPFind gets bad control points for stacks

[tduell]

Working with hugin_executor branch 6769:aed7abb78308, built on Fedora 20 x86_64.
This issue has been difficult to reproduce.
I have 3 sets of bracketed images, each 3 exposures.
Hugin detects the stacks and I ask for positions to be linked.
CPFind finds control points for images 0-3, 0-6, and 6-3, as expected, but also finds 23 control points for 0-0.
The group of CPs for 0-0 has the same pattern in each view of image 0 in CP tab, but the group has been displaced.
I have attached a screenshot of the CP tab.
At this stage I haven't been able to reproduce this using a set of downscaled images, so providing the project images is a bit moot at this point.

[tduell (tduell-iinet)]

Here is the .pto file

[tduell (tduell-iinet)]

Finally managed to get the problem to show up using downscaled images.
Attached zip archive has the small images and the .pto exhibiting the described behaviour.

Cheers,
Terry

[tmodes (tmodes)]

Should be fixed in repository in changeset 7a2a12e71e67.
Thanks for the test case. It help to find the bug.

[tduell (tduell-iinet)]

The problem hasn't arisen again in the testing I've done thus far.
Thanks Thomas, you're worth your weight in cocky's feathers :-)

Cheers,
Terry