support state override for debug_traceTransaction ("offchain events")

[hananbeer]

the debug_traceCall function supports state overrides, but debug_traceTransaction does not - it only accepts a transaction hash.

I wrote this originally as the backbone for generating "offchain events" (as developed by shadow / ghost logs)

to prevent confusion with nodes that don't implement this feature I created a distinct function debug_traceTransactionOverrides.
by overriding contracts bytecode it is possible to reexecute onchain transactions in different context, specifically to emit so called "offchain events" but not limited to this.

see equivalent implementations:

reth: paradigmxyz/reth#7410
erigon: erigontech/erigon#9847

[s1na]

Why is this not possible by using debug_traceCall and passing in all the tx args? traceCall has the option to produce the state of a (block, tx_id).

[hananbeer]

Why is this not possible by using debug_traceCall and passing in all the tx args? traceCall has the option to produce the state of a (block, tx_id).

I haven't seen this in docs.
https://geth.ethereum.org/docs/interacting-with-geth/rpc/ns-debug

do you have reference?

[s1na]

I haven't seen this in docs.

Sorry about that. It seems we forgot to update the docs after adding it. This is the PR #28460 in question. So it's added to the last object which has the state and block overrides and tracing config.`

@jsvisa care to update the docs?

[jsvisa]

I haven't seen this in docs.

Sorry about that. It seems we forgot to update the docs after adding it. This is the PR #28460 in question. So it's added to the last object which has the state and block overrides and tracing config.`

@jsvisa care to update the docs?

Sure, I'll add it in the docs later today or tomorrow

[jsvisa]

@s1na how about let's also add TxIndex for the eth_call RPC, seems that is also a useful feature, wdyt.

[s1na]

@s1na how about let's also add TxIndex for the eth_call RPC, seems that is also a useful feature, wdyt.

@jsvisa Yep I think it makes sense. But with multicall around the corner I'd say it's a short term benefit anyway.

@hananbeer let me know if you tried the alternative I suggested and it works for you, then I'd go ahead and close this PR. We can keep the debug_traceTransaction a pure re-execution while adding allowing all override possibilities in traceCall.

[hananbeer]

that solution is fine I just want it to be standardized across nodes :)
I guess we can close this pr.

[ArtificialPB]

IMO, it still makes sense to support state overrides on both TraceTransaction and TraceBlock calls. This will allow tracing a whole block with injected events, with a single call, instead of doing a TraceCall for each tx in the block.

It would be a two-line code change: just move StateOverrides *ethapi.StateOverride from TraceCallConfig to TraceConfig.