Seeding (DNS and fixed) on regtest #5397

Regtest has neither DNS nor fixed seeds. The net.cpp 1229-1237 block in ThreadOpenConnections() acts oblivious to this, and attempts to add fixed nodes, assuming DNS seeds are down. While this may not necessarily be entirely inappropriate (in terms of 'path invariants'), this also may not remain harmless indefinitely, if we are assuming the possibility of an 'infrastructure attack' as the comment does (legitimate in the context of mainnet, or even testnet). At the very least, the comment itself needs to be updated to call out the regtest use case.

What attack would you imagine on this? regtest is - by definition - local-only so any type of DNS seeding or fixed seeding would be actively harmful. You could add a check that those stages are skipped entirely iff there are no seeds, but in practice nothing would change?

I agree completely. The concern would be any future mods to this block, should they overlook the regtest use case (and therefore possibly take more dramatic measures operating under the assumption of an infrastructure attack). So, what say you? Is a comment update in order, or a comment update + skip the block entirely if there are no seeds? I have to say the code has managed to handle the different modes remarkably uniformly so far...