I think you may want to use Equals() rather than an exact floating point comparison. Search for it to see usage.

Equals only seems to be for Vectors, I looked at it's implementation and it uses LENGTH_EPS, does this make sense to use for the Theta though?

I updated the code to also remove vertical and horizontal constraints when not rotating 90, 180, 270, 360 etc degrees, otherwise the drawing would jump around.

I also used fmod to make the automatic swapping of vertical and horizontal constraints more generic.

Just found in the import code (importidf.cpp) that it uses 0.01 for the theta. will look into this some more...

Maybe EXACT is what @phkahler had in mind. It is a macro that disables compiler warnings for using == on floating poits - which normally is stupid.
So

if (fmod(theta + (PI/2), PI) == 0)
Becomes
if (EXACT(fmod(theta + (PI/2), PI) == 0))

Alternatively:
if (fmod(theta + (PI/2), PI) < LENGTH_EPS)

an fabs may be needed - I'm not sure... writing from the phone...

Negative numbers should also work with fmod, so I think the fabs in unnecessary

I just meant there should be some tolerance around the angle. What if someone types 3.14159/2 or if FP comparison isnt quite exact for whatever reason.

I dont remember if we have an FP version of wrap(). Cant look right now...

Actually, the rotation is provided in degrees by the user, if someone types 90.00001, that still means we don't want to swap the V/H constraints, they should be removed because they would not be perfectly horizontal or vertical anymore.

I think the EXACT makes the most sense here.

In my opinion this you should have squashed this PR into one commit.

There is a option in the repository settings to disable merge commits / only allow squash which might be helpful to enforce this for future PR's

@vespakoen no, because a PR may contain multiple commits that make sense to be separate. In fact I think most (all) PR-s should be merged with rebase.

One more note - you made the swap-vertical-horizontal-constraints on the solvespace/solvespace (https://github.com/solvespace/solvespace/tree/swap-vertical-horizontal-constraints) upstream repository instead of your own fork. I am not sure (I don't have much experience with contributing to open source) but think that this is considered bad form.

Ah shoot, thanks for letting me know, this was not my intention. I should really be more careful with my remotes, my "muscle memory" is used to working with the main repository directly, which probably caused this issue...