Incorrect feedrate with six axes?

I don't wish to discourage you but the guy who added the lookahead code has said that adding ABC rotary axes is problem riddled.
Adding UVW might be easier since they are linear.
IIRC, there are some google videos writeups in machinekit that talk of how it works.

github.com/LinuxCNC/linuxcnc/tree/master/src/emc/tp

There was an experimental branch that extended the new tool planner to UVW.

That sounds promising! Can you point me in the direction of that branch?

This may be a little old (it pre dates the JA merge into Master).
github.com/LinuxCNC/linuxcnc/tree/feature/uvw-blending-dev

That branch was started partially at my behest. I needed blending to work on a machine with XYZW axis (two Z axis). By the time this was ready for testing I had already figured out another work-around that worked for my application. So the work of maintaining a git-pull on a factory production machine did not appeal to me, so I never used it beyond some simulations and few test runs. (That is what I remember from 3 years ago and the accuracy of my recollection could be suspect.)

Thanks for the pointers! I've already downloaded the code and started poking through it. I actually need this code to build on machinekit since I'm running a beaglebone, but I (correctly) assumed that the trajectory planning code is mostly shared. If I can get things to work correctly, I'll think about getting a PR together.

Just curious, what was your solution in your XYZW machine?

I set up a way to temporarily slave/unslave the W to the Z axis using some custom M-codes and a bunch of Hal logic trickery.

If you have two routers why not run each with a separate Computer and sync them in G code with I/O ?.

Joe, unfortunately the two machines are sharing some components, and hence can run into each other. They will need to react to each other pretty intimately to avoid damage. I considered trying to set up a separate tp instance to control the two halves independently, but was afraid of the interference.

Todd, good news! I got your old code working in Machinekit, and it's a dream! Initial tests show that everything is behaving correctly. I've got both sides of the machine running at the full indicated feedrate. Thanks again for your help, and your old code! I might try to open a PR for machinekit, but I have a suspicion they might want a cleaner implementation. I see stubs in their codebase for full 9-axis blending.

Glad you were able to find something that works for you.