…18988)

Problematic scenario encountered while working on
erigontech#18290

When requesting proof of a non-existent storage key for an account that
has no storage, the traversal does not stop after the account is
encountered, and garbage data from the rows in the grid below the
account are incorporated into the proof. This ultimately results in
incorrect trie root hash.

This problem never surfaced with `eth_getProof` calls because that RPC
call has logic to avoid calling `GenerateWitness` when the account in
the account proof has an empty storage root.

The fix is to stop the traversal if `len(hashedKey)>64` and storage root
of account is empty.

---------

Co-authored-by: antonis19 <antonis19@users.noreply.github.com>

…18988)

Problematic scenario encountered while working on
erigontech#18290

When requesting proof of a non-existent storage key for an account that
has no storage, the traversal does not stop after the account is
encountered, and garbage data from the rows in the grid below the
account are incorporated into the proof. This ultimately results in
incorrect trie root hash.

This problem never surfaced with `eth_getProof` calls because that RPC
call has logic to avoid calling `GenerateWitness` when the account in
the account proof has an empty storage root.

The fix is to stop the traversal if `len(hashedKey)>64` and storage root
of account is empty.

---------

Co-authored-by: antonis19 <antonis19@users.noreply.github.com>

Some tests in `Test_WitnessTrie_GenerateWitness` incorrectly passed
because the cached trie node hashes were cached, but when `tr.Reset()`
was being called, the unit tests started failing which exposed issues in
the trie structure.

Due to this, problems appeared with the witness of
#18290 because proofs are
requested even for storage keys that haven't been created yet for
accounts with no storage, and there was no logic to prevent traversing
further than the account. There was a handling introduced for that in
#18988 but the fix there was
insufficient and the caching hid that the problem wasn't fixed.

This PR fixes the unit tests correctly.

---------

Co-authored-by: antonis19 <antonis19@users.noreply.github.com>