Fix genesis blocks for testnet, testnet4, signet and regtest

[argfamobile]

Problem

-regtest, -testnet and -signet cannot start at all. On a clean datadir the node writes the genesis block and then fails to read it back:

init message: Verifying blocks…
[error] Errors in block header at FlatFilePos(nFile=0, nPos=8) while reading block
[error] A fatal internal error occurred: Failed to read block.
[error] Failed to connect best block (Failed to read block.)

Reproduced with the official btgs-x86_64-linux-gnu.tar.gz release of v30.2.3, and with two independently compiled builds (GCC 12 on Debian 12 and GCC 15 on Ubuntu 26.04), so it is not specific to any toolchain.

Cause

The genesis parameters for the test networks were still Bitcoin's, while BTGS uses its own timestamp message:

"The Times 31/Mar/2026 BTGSG Chancellor on brink of second bailout for banks"

A different message means a different coinbase, a different merkle root and therefore a different block hash. Bitcoin's nonces no longer satisfy the proof of work for these chains. The hashes they actually produce are:

network	nBits	old nonce	resulting hash	meets PoW
testnet	0x1d00ffff	414098458	4d7af693712b8ff7...	no
signet	0x1e0377ae	52613770	2d63e799bcb90364...	no
regtest	0x207fffff	2	f612acbdbffba76f...	no

testnet4 did start, but only because it kept Bitcoin's timestamp message and output script verbatim, so its genesis was literally Bitcoin's testnet4 genesis rather than a BTGS one.

The eight hashGenesisBlock and hashMerkleRoot asserts were commented out in these four networks. Those asserts are exactly what would have caught this on the first run. Mainnet has them active, and mainnet is correct. With them disabled the node accepted a genesis block that does not meet its own proof of work, and the failure only surfaced later as a block read error, which points at the wrong part of the code.

Fix

Mined new nonces using the BTGS genesis message, so each chain gets its own genesis block:

network	nTime	nonce	genesis hash
testnet	1296688603	640026520	00000000dec2f478c7277b937818c3ad7515ce9ca7c9a5f8f6345e5c18f78381
testnet4	1714777861	453240970	000000001d07bb1394fd589dbd041c1618cde63ba6d611e7b1f0c82678f58697
signet	1598918400	8705482	000002e46f9760f1d878dab0735f04a592c8d4bcdb6bfcd9c7b321f54e2e8ba7
regtest	1296688602	0	7bb66a7fba8df679c927b239aa09d5f40ce6600eb8aea11b915bb7844517969a

Merkle root for all four: 9349824da96d7c3c2a9e8a5ab54b61e14df2c77788b0b25aaa178545362bebc8

testnet4 now uses the standard CreateGenesisBlock overload with the BTGS message, replacing the inherited Bitcoin message and script.

All eight asserts are re-enabled. If a genesis is ever wrong again, the node will refuse to start instead of starting broken.

Mainnet is not touched. Its genesis, its nonce and its asserts are unchanged.

Verification

Built from this commit and started the node on all five networks:

[mainnet ] OK  0000000d1c5a497963a46c0348cb4346779c52d9e1d7cc8b5efb1be0a4a0f964
[testnet ] OK  00000000dec2f478c7277b937818c3ad7515ce9ca7c9a5f8f6345e5c18f78381
[testnet4] OK  000000001d07bb1394fd589dbd041c1618cde63ba6d611e7b1f0c82678f58697
[signet  ] OK  000002e46f9760f1d878dab0735f04a592c8d4bcdb6bfcd9c7b321f54e2e8ba7
[regtest ] OK  7bb66a7fba8df679c927b239aa09d5f40ce6600eb8aea11b915bb7844517969a

Since the asserts are runtime checks, the fact that all five networks start is itself the proof that the values are correct. A wrong genesis would abort at startup.

Also created a chain from scratch on regtest and mined 101 blocks:

altura inicial : 0
address        : bcrtg1qrd43lrxufurejwssqp27u54pzsvr3tnuulmvmg
tras minar 101 : 101
balance        : 2500000.00000000 BTGS

That is what makes test/functional/ usable. The 204 functional tests in the repo all run on regtest, so none of them can execute today.

Notes

Branched from the v30.2.3 tag as requested. v30.2.3 is an ancestor of main, so this applies cleanly.

Reported and discussed with @Annonone on Discord before opening.

Happy to adjust anything, and glad to test further changes on all networks.