I was testing non-standard signature hash flags on test and regtest and was getting these rejections: "non-mandatory-script-verify-flag (Signature hash type missing or not understood)"
This change to AcceptToMemoryPoolWorker simply allows the mempool of nodes on the test and regtest networks to accept transactions containing signatures with non-standard signature hash types.
1351 | @@ -1352,7 +1352,7 @@ bool AcceptToMemoryPoolWorker(CTxMemPool& pool, CValidationState& state, const C 1352 | 1353 | // Check against previous transactions 1354 | // This is done last to help prevent CPU exhaustion denial-of-service attacks. 1355 | - if (!CheckInputs(tx, state, view, true, STANDARD_SCRIPT_VERIFY_FLAGS, true)) 1356 | + if (fRequireStandard && !CheckInputs(tx, state, view, true, STANDARD_SCRIPT_VERIFY_FLAGS, true))
I'm fairly sure that this is not correct. You certainly shouldn't be skipping the inputs check when standard-ness is not required, it is part of consensus not standardness.
There is a second invocation to CheckInputs below, with just mandatory checks.
It may make sense to disable the extra validation flags in some case for testing, but I don't think it should be done unconditionally.
The standardness validation flags don't really disable any functionality (that's done through IsStandard), but some are for example used to roll out softforks safely. Disabling those entirely on testnet means that it can't be used to test whether such relay rules actually work.
It's still changeable per node with "-acceptnonstdtxn" on test and regtest.
The way I was looking at this commit was that it makes the application or non-application of standard-ness validation, as controlled by "-acceptnonstdtxn", more complete.
I think fRequireStandard in the chain params is false for these two networks in order to not inhibit propagation and inclusion of possibly surprising valid stuff, and hopefully catch related bugs in non-prod testing.
I see your point though too. On regtest this change was definitely helpful for my local testing. On testnet3, because standard-ness validation is the default for what I was testing in current nodes, it's kind of sterile. I was having trouble getting relay and I couldn't find good info on current testnet miner policies or where I could submit to a miner that accepted non-standard testnet transactions (which makes sense, but it was kind of a pain).
Doesn't look like this is gaining any traction. I'll close it in a few days if nobody objects.