tracers: tx-level hooks

[s1na]

We have EVM-level hooks (CaptureStart, CaptureEnd) and call-level hooks (CaptureEnter, CaptureExit). I'm proposing adding something like CaptureTxStart and CaptureTxEnd.

Why should this feature exist?

This feature makes the following information easily available to tracers: start gas (or intrinsic cost), refund values, any pre-check errors and if we want tx fields.

edit: actually it's not possible to accurately compute the intrinsic cost now in tracers because they don't have access to the access list.

What are the use-cases?

Main reason I started thinking about this is in some of the tracers we have to compute the intrinsic cost & refund value (e.g. in the logger).

Implementation

Invoke the hooks here:

go-ethereum/core/state_transition.go

Line 275 in 8fddf27

func (st *StateTransition) TransitionDb() (*ExecutionResult, error) {

[rjl493456442]

Is it possible to merge the CaptureStart and CaptureEnter for both call frame scope? Looks like the only difference right now is CaptureStart is for topmost call frame?

Or we keep it for backward compatibility concern?

[s1na]

Is it possible to merge the CaptureStart and CaptureEnter for both call frame scope? Looks like the only difference right now is CaptureStart is for topmost call frame?

I agree it makes sense semantically to have CaptureEnter for all (including) topmost call frames, and move up captureStart to emit only tx-level relevant info. I think it'd need bigger changes in our tracers but I'm gonna give it some more thought.