This doesn't properly handle attempting to constrain the entities tangent a second time if their remaining endpoints are not already coincident. It redundantly constrains the first two points.

This also cheats when it detects the existing constraint. It rewrites that first constraint to constrain the other points and then proceeds to constrain the first two points again. It works but it might cause a problem somewhere else, changing the constraint after it's been written.

These can be fixed but the code started to get even uglier. This is still a work in progress.

@ruevs Didn't mean to conflict with your commit. You might be in a better position to fix this since I think you are working on the tangent #937. It's tricky though. There's a couple of frustrating edge cases.

No problem. I need time to think about this and debug through it... corner cases...
This may be for after 3.0 - especially considering that it can be done with construction entities.

See here #951 (comment)

I rabased this on master here and it works.

It has an interesting side effect when a curve and line do not share end points - if the user selects the two entities and one end point of the curve and applies the tangent constraint it works as expected. If he repeats exactly the same a second time the other end of the curve gets constrained. Not surprising and perhaps not a big problem.

But see here #951 (comment) in current master it is possible to achieve this without this PR by selecting the entities in a different order.

Hmmmm...