Issue with Inverse Kinematics Failing During Homing – 3-Axis Robot with LinuxCNC

Note that all that I know about kinematics is that it confuses me...

Forward kinematics takes the joint positions and calculates the end effector position. This will always work and will always return exactly one solution.

Inverse kins takes the required end effector position and calculates the joint positions to achieve that. Sometimes there is no solution (because the arm simply can't reach that point). Sometimes there are multiple solutions (in fact there _generaly_ are.
So, the kins has to choose a solution. And sometimes it can lock itself in a corner, where there are solutions for the current position, and for the next position, but it would have to pass through a "wrong" position to get there.
This isn't just robots. I am sure you have been in this position with your own arms.... The difference is that you have a brain and are not blindly following what is basically a relatively simple mathematical equation which has no predictive capability.

So, back to your questiion: Homing happens in joint space, using forward kinematics. So presumably you only see the error after homing is completed, and the system switches to "world mode" and inverse kins.
Maybe your home position (in XYZABC rather than joint angles) simply isn't reachable? (or, doesn't _actually_ correspond to the joints-homed position?)

Do the joints not move at all when you try to home them?

As a short-term expedient you can turn off the requirement to home (INI NO_FORCE_HOMING) and (if you are using the Axis interface) switch to world mode with Ctrl-$.

But don't forget that then the machine will have no idea where the joints are, so the end-effector motion will be a very curly version of cartesian.

Note that 'genserkins' uses 'modified' DH parameters.
Have a look here (there is also an example of how to derive the parameters for a manipulator):
www.linuxcnc.org/docs/stable/html/motion...ml#cha:dh-parameters