@jeffdi @Tim Edwards Can it be, that in pad

sky130_ef_io_gpiov2_pad_wrapped

the wire

PAD_A_NOESD_H

is not properly connected to the pad? This pad connection is also exposed in the analog caravel (as

gpio_noesd

). Without proper documentation of the pad (which I don’t find) and without the nestlist of the pad (which I don’t have) it is impossible to confirm, but looking at the magic file there seems to be a small gap (see screenshots). Can someone from efabless please confirm?

I think the "gap" you're looking at is the metal3 resistor that separates the "PAD" net from the "PAD_A_NOESD_H" net.

Thanks @Tim Edwards, that explains it. I did not think of a metal-res 😞 How about adapting the tracing feature in magic, the

s-s-s

that it also jumps across metal-res and show it as connected? From a logical sense I think this would be the right behaviour, as metal-res are mainly there for net name changes.

@Harald Pretl the xschem schematics are here.

Thanks @Mitch Bailey good to know!

@Harald Pretl: That's probably doable but requires telling magic that it has to modify the connectivity tables. Currently it has no knowledge that a metal resistor is any different from any other type of resistor.