Someone logged a bug against RioRand_LM2596 ("a problem pcb lm2596 #3282") from core on the github bug tracker. Allthough for me, while search finds it fine, I can't, so I don't know exactly which bin its in! Since I have one of these (or more correctly a clone from Ebay) and thus have an interest in a working Fritz part, I had a look and indeed the part is broken. The PCB view doesn't match the board (possibly because the intent was to use wires to the board), but more important is that breadboard view is very broken, only pins 3 and 4 actually connect in breadboard. So having the part in hand to measure I fixed both these problems in varient2 (varient1 provides the original broken implementation if selected in Inspector without me knowingly doing anything. Nice feature!). The breadboard view had pins 1, 2 and 4 set as bendable legs, but without a connection outside the viewbox which is why 1 and 2 don't work in the original. I'm not sure why 4 worked but it seems to. As usual correcting the xml in Inkscape was a learning experience but I have found a new way to get nodes in to groups so thats a plus. On the PCB side, I changed the layout to match the physical module and added holes for the two mounting posts (so you could mount a module to a board on standoffs and connect either wires or pins between the board and the module) and deleted the keepout layer. Testing with 4 single pin connectors (one for each input and output) now works in all views and the PCB should now match reality (complete with pins with .05 offsets from a .1 grid just like the actual board I have) but I haven't actually cut a board from it to verify that the mounting holes and pins are in fact in the correct place. In any case here is the new part:
(Edit: updated part to correct errors in pcb view)
RioRand_LM2596.fzpz (42.4 KB)
I'll add a pointer to this upload to the github problem report in case someone is actually looking at them ...
Peter Van Epp