Gas control tweaks

[edfelten]

This makes some changes to simplify the semantics of gas, and eliminate some non-intuitive behaviors.

First, when a sub-tx finishes, it verifies that the sub-tx's unused gas would be enough to pay for the net storage allocation done by the sub-tx. If not, it reverts the sub-tx, as if it had run out of gas. This ensures the invariant that a sub-tx can't deplete its parent's gas by more than the gas amount given to the sub-tx.

Second, the EVM GAS instruction now fully accounts for the storage cost that the current tx would incur if it finished immediately. If the current tx has been a net allocator of storage, the number of cells allocated, multiplied by the current ArbGas cost per storage cell, is subtracted from the remaining gas, to determine the value returned by the GAS instruction. If the result would be negative, the tx is immediately treated as out-of-gas. This causes the GAS instruction to have more intuitive behavior.

[codecov-commenter]

Codecov Report

Merging #552 (99c0c71) into london-hardfork (a38f1b5) will decrease coverage by 0.11%.
The diff coverage is 82.69%.

@@                 Coverage Diff                 @@
##           london-hardfork     #552      +/-   ##
===================================================
- Coverage            76.85%   76.73%   -0.12%     
===================================================
  Files                   84       84              
  Lines                29358    29395      +37     
===================================================
- Hits                 22562    22557       -5     
- Misses                6796     6838      +42     

Impacted Files	Coverage Δ
arb_os/evmCallStack.mini	83.11% <69.23%> (-0.22%)	⬇️
arb_os/evmOps.mini	74.43% <72.72%> (-0.08%)	⬇️
arb_os/gasAccounting.mini	81.64% <92.30%> (-0.08%)	⬇️
arb_os/chainParameters.mini	80.48% <100.00%> (ø)
arb_os/retrybuffer.mini	81.18% <100.00%> (ø)
src/compile/typecheck.rs	64.66% <0.00%> (-0.98%)	⬇️
src/run/emulator.rs	53.90% <0.00%> (-0.06%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a38f1b5...99c0c71. Read the comment docs.

[rachel-bousfield]

Our gas-charging model has evolved over time via other PRs, so this one is out of date with our current thinking