run_block_generator2()

[arvidn]

This PR is best reviewed one commit at a time.

This updates the hard-fork to use run_block_generator2() as intended. this alters the cost block generators as the cost of the ROM itself is not charged (once the hard fork activates)

The new run_block_generator2() can be reviewed here: https://github.com/Chia-Network/chia_rs/blob/main/src/gen/run_block_generator.rs#L111-L191

Since testnet10 has already activated the hard-fork, this fix needs a separate activation height there. In mainnet, the activation stays the same as the original hard fork.

There are new pre-generated test blockchains in the test-cache repository, with the _harfork suffix.

[coveralls-official]

Pull Request Test Coverage Report for Build 6122871321

• 66 of 67 (98.51%) changed or added relevant lines in 9 files are covered.
• 8 unchanged lines in 3 files lost coverage.
• Overall coverage increased (+0.01%) to 89.139%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
tests/core/mempool/test_mempool.py	9	10	90.0%

Files with Coverage Reduction	New Missed Lines	%
chia/timelord/timelord.py	1	80.39%
chia/wallet/wallet_node.py	1	87.09%
chia/full_node/full_node.py	6	85.96%

Totals
Change from base Build 6118026788:	0.01%
Covered Lines:	86477
Relevant Lines:	96958

---

💛 - Coveralls

[arvidn]

I'm removing coverage-diff label. the one missing line points into a benchmark, which we don't instrument for coverage

[richardkiss]

I don't quite understand the intricacies of all the new test branches, although I do have the idea that they have to do with different costs for the new run_block_generator2. I assume the new costs are always lower than the old costs, which will give us a slightly simpler hard fork discontinuity (non-smoothness) of expecting the committed costs to be below the actual block_generator2 costs (maybe even by a constant?).

[arvidn]

I assume the new costs are always lower than the old costs, which will give us a slightly simpler hard fork discontinuity (non-smoothness) of expecting the committed costs to be below the actual block_generator2 costs

correct.

(maybe even by a constant?)

it won't be constant. the most expensive thing the generator ROM does is to compute the puzzle hashes of puzzles. This will be free after the hard-fork, so the "savings" depend on how expensive the tree-hashes are, essentially.

[github-actions]

File	Coverage	Missing Lines
tests/core/mempool/test_mempool.py	90.0%	lines 2493

Total	Missing	Coverage
73 lines	Unknown	98%

[wjblanke]

aok thanks guys!