While debugging #540, I realized that I was running cargo test from two different clones of omicron with the same value of $TMPDIR in my environment, and that they could totally stomp on each other if one of their "build" steps initialized $TMPDIR/crdb-base while the other one's test suite was running trying to create a new database directory from that directory.
If we move the seed directory to ./target, it will at least be scoped to one local clone of Omicron.
While debugging #540, I realized that I was running
cargo testfrom two different clones ofomicronwith the same value of$TMPDIRin my environment, and that they could totally stomp on each other if one of their "build" steps initialized $TMPDIR/crdb-base while the other one's test suite was running trying to create a new database directory from that directory.If we move the seed directory to
./target, it will at least be scoped to one local clone of Omicron.