Open and auditable source code is a laudable goal, and one I generally endorse.
But the more important issue is an open audit trail.
The implementations I’ve seen that make the most sense are electronic machines that validate and mark ballots that are both human readable and machine readable. The input validation can prevent overvotes (accidentally voting for more than one candidate) and add a verification step for undervotes (choosing to leave a particular choice blank), while the voter gets a verifiable visual feedback that their ballot has been properly created. Then they drop it in the box.
At the end of the night, the paper ballots are fed into tallying/counting systems, which should entirely distinct from the input validation systems. That way they get a machine count that night, but still have an auditable paper trail.
Given the choice between a direct voting machine that’s open source, or a closed source machine that creates the paper trail in that way, I’d choose the auditable process.