client: Fix double runs of the block execution

[g11tech]

On the 4844 devnet5 a weird error was noticed:

1. on new payload the client runs vmExecution.runWithoutSetBlock first and
   
2. then on fcU once the block is part of canonical chain it runs vmExecution.run which runs block in an iterator loop
   

Run on 1. gives VALID execution while 2nd fails with invalid state root

This PR aims to solve it:
Synopsis of the issue: the vmExecution's run and runWithoutSetHead can end up running the same block one after another leading to state manager's incorrect caches

This PR updates clearCache flag in runBlock if such a scenario comes up during client execution.

• Add a testcase to repro
• Fix the issue
• Testing
  • sync mainnet
  • sync local post merge devnet
  • beacon sync and stay sync in a devnet

PS:
the issue is resolved with this diff which might give an idea about where the problem could be i.e. by running the block again in the vm copy of vmexecution

--- a/packages/client/src/execution/vmexecution.ts
+++ b/packages/client/src/execution/vmexecution.ts
@@ -353,7 +353,8 @@ export class VMExecution extends Execution {
                 }
                 const beforeTS = Date.now()
                 this.stats(this.vm)
-                const result = await this.vm.runBlock({
+                const ivm = await this.vm.copy()
+                const result = await ivm.runBlock({
                   block,
                   root: parentState,
                   clearCache: reorg ? true : false,

[codecov]

Codecov Report

Merging #2730 (879cbad) into master (9c8f407) will decrease coverage by 0.31%.
The diff coverage is 62.88%.

Additional details and impacted files

Flag	Coverage Δ
block	90.82% <ø> (ø)
blockchain	90.72% <ø> (ø)
client	86.82% <63.21%> (+<0.01%)	⬆️
common	97.05% <ø> (ø)
devp2p	?
ethash	∅ <ø> (∅)
evm	79.58% <ø> (ø)
rlp	∅ <ø> (∅)
statemanager	80.53% <ø> (ø)
trie	89.92% <ø> (-0.34%)	⬇️
tx	95.76% <ø> (ø)
util	81.13% <ø> (ø)
vm	81.36% <0.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

[g11tech]

a lot of diff is because of the formatting changes by moving runWithLock a bit out, but the real PR work is till this commit: 810aa54

[holgerd77]

Rebased this via UI

[holgerd77]

Lot of great smallish additions/clean-ups and nice partial VM execution refactoring, this is still very much needed to have improvements there!

Code looks good, and I guess execution-flow-code can also best be tested by the test cases we have + continued client runs, where I am at right now anyhow.

So will approve and merge. 🙂

[holgerd77]

👍 🙏

[holgerd77]

Nice new test case, VM set iterator head execution tests very much needed, great! 👍