open-source-silicon.dev

does anyone know what "Missing terminal connection of device at &lt;location&gt;" means when doing circuit extraction in Magic?

in this project: <https://github.com/mattvenn/tt06-analog-r2r-dac/tree/main/mag>

running: `PROJECT_NAME=r2r make r2r.sim.spice`

works on my  PC, but causes a bunch of warnings on another PC

could be to do with a different PDK version

```Port: name = b7 exists, forcing drivepoint
Location is (3000, 17400); drivepoint (3000, 17400)
Port: name = b0 exists, forcing drivepoint
Location is (-1199, 17400); drivepoint (-1200, 17400)
Port: name = GND exists, forcing drivepoint
Location is (-2599, 2600); drivepoint (-2600, 2600)
Port: name = VSUBS exists, forcing drivepoint
Location is (-2585, 3016); drivepoint (-2586, 3016)
Missing terminal connection of device at (1366 8398) on net b4
Missing terminal connection of device at (166 8398) on net b2
Missing terminal connection of device at (-1834 2998) on net GND
Missing terminal connection of device at (-434 8398) on net b1
Error: Node with no area.
Missing terminal connection of device at (1966 8398) on net b5
Missing terminal connection of device at (766 8398) on net b3
Error: Node with no area.
Missing terminal connection of device at (2566 8398) on net b6
Error: Node with no area.
Missing terminal connection of device at (-1034 8398) on net b0
Error: Node with no area.
Error: Node with no area.
Error: Node with no area.
Error: Node with no area.
Missing terminal connection of device at (3166 8398) on net out
Missing terminal connection of device at (2566 2998) on net out
Missing terminal connection of device at (3166 8398) on net b7
Total Nets: 45
Nets extracted: 18 (0.400000)
Nets output: 11 (0.244444)
exttospice finished.
rm -f *.ext```


<@U0172QZ342D> I think this occurs most often on decap cells. The mosfet devices are expected to connect to 4 terminals, but the source and drain connect to the same node. You can save the feedback file and look at each warning/error.
```feedback save ext_feedback.out```

Sometimes it indicates a real problem that magic has with some device type.  The result of the problem is that the terminal it can't find defaults to the root name of the net being segmented, so in effect it moves the device terminal to the root of the net (as in, right next to the pin) so the delays to that device will be wrong.  The impact on simulation results is likely to be minimal.  But it's one of those things I'll want to dig down into later and figure out how to fix.