p2p: isolate regtest from mainnet bootstrap state

[jeetrex17]

Fixes #10392.

--regtest wasn't being handled in the p2p layer, so it defaulted to mainnet. That meant regtest could resolve seeds.moneroseeds.*, fall back to the hardcoded mainnet seed list, and share p2pstate.bin with mainnet.

This patch makes p2p treat --regtest as FAKECHAIN, skips automatic public seed bootstrap for that mode, and stores p2p state under <data-dir>/fake to keep it separate from mainnet.

[nahuhh]

This doesnt work as expected.

example:

when using --testnet the default blockchain folder will be ~/.bitmonero/testnet/lmdb

when using --regrest the default blockchain folder will be ~/.bitmonero/fake/lmdb

when NOT using the default folder (ie, when using somethinf like --data-dir testing the latter will fail with an error of:

2026-04-02 21:48:38.803 E Exception in main! boost::filesystem::create_directory: Not a directory: "testing/fake"

This error does not happen with --testnet. Using data-dir with testnet will create a folder like ./testing/testnet

Can you make it create the subfolder in the same way that testnet and stagenet subfolders are created?

(Also, i would prefer if the subfolder be called regtest instead of fake)

[jeetrex17]

Thanks for pointing that out. Addressed in 3348a4b.

Regtest data-dir handling is now normalized through arg_data_dir the same way as testnet/stagenet, so both core and p2p use <data-dir>/regtest instead of appending a subfolder later. I also switched the regtest subfolder from fake to regtest.

[j-berman]

Also, i would prefer if the subfolder be called regtest instead of fake

Considering the subfolder for the blockchain dir is already located in fake when using --regtest, I'd advocate for continuing to use fake to avoid breaking any flows that may rely on current behavior. There is even explicit logic for removing or reusing the already stored data in that dir

[nahuhh]

the subfolder for the blockchain dir is already located in fake when using --regtest

you're right. I neglected to pay attention to what the current behavior was

[jeetrex17]

Makes sense, thanks. Switched it back to fake in 7a14dea while keeping the centralized data-dir handling.

[nahuhh]

Ran a few different tests (generating blocks, peering, syncing wallet, sending txs etc). this pr appears to resolve the linked issue, and is working as expected / havent found any other regressions (after data-dir was fixed)