Forwarding this issue from the TT discord as it might have more eyes here. The .mag file at https://github.com/hugodiasg/tt06-temp-sensor/blob/main/mag/device_without_rf.mag looks DRC clean, but exporting it to .gds and reading back the .gds gives a few dozen DRC errors. What's going on?

@htamas can you include the magic and tech file (open_pdks) version?

I've used the versions in the iic-osic-tools docker image, i.e. magic 8.3.416 & open_pdks cd1748bb197f9b7af62a54507de6624e30363943. Matt reproduced it with magic 8.3.447.

@htamas: Okay, the problem here is that the layout .mag file was flattened, which I keep telling people not to do, and this is one of the reasons why. There are slightly different DRC rules for devices in standard cells; SkyWater (or rather Cypress, originally) bent the rules a bit for the standard cells and then made up a few "rule exceptions" in the rule deck for them. The DRC engine, whether it's magic, klayout, or Calibre, relies on the standard cell ID layer to determine that the DRC rule exception applies. The problem is that by flattening the circuit, magic no longer knows where the standard cell boundary is, and it doesn't get into the GDS output. All those DRC errors suddenly pop up when the GDS is read back. So this design would fail tape-out precheck. So again I repeat: Do flatten circuits for obtaining an R-C parasitic extracted netlist. Don't save them in the flattened state.

Thanks. We'll keep an eye on this for future analog TT submissions.