It might be safer to use this: https://github.com/efabless/openlane/tree/master/designs#update-design-configs-for-a-pdkstd_cell_library-pair-after-an-exploration Or mimic its behavior

Also, I'd advise updating openlane to

master

and using the rc5 tag.

git pull
git checkout rc5

and then rebuild. Magic had an issue with flagging false metal hole violations that was fixed sometime last week, so updating openlane to latest rc5 would include that magic version, just in case these spikes you're seeing are false errors.

any clues on what --pdk or --standard-cell-library should be set to?

you said: They are, which file are you copying exactly?

I'm just iterating over fp_core_util, then opening the csv file, finding one with no drc, and copying that one number into my config.tcl

Then this is probably the magic issue I mentioned above

Try updating openlane

ok

I'll update

when you say rebuild

should I run make in the top level to rebuild the pdk too?

or just make merge in ./docker_build

However, things should now be stable since we moved to master of openlane and fixed a commit for open_pdks and skywater-pdk, (so rebuilding the pdk would be a good thing to do). However, let's try to know the reason for the strange behavior you're getting first and then we can update the pdk.

thanks