stateManager.setStateRoot doesn't clear the _storageTries cache.

[davidmurdoch]

#417 was about the _cache which was fixed in #420. This issue is about _storageTries https://github.com/ethereumjs/ethereumjs-vm/blob/c842d101108d6acdd9ca82ab1de274ba3b16e5b2/lib/state/stateManager.js#L38

which I think should also get cleared when the stateRoot is set.

We ran into the issue with Ganache when a user creates a snapshot (which just means we record the block number at the time snapshot was called), then modifies contract storage, reverts the snapshot (we just set the state root back to the way it was when the snapshot was created), then modifies contract storage again.

PR and tests to follow.

[s1na]

What you're describing seems to be really similar to what commit and revert in stateManager are supposed to do. Have you checked them out by any chance? revert does clear both the state cache and the contract storage cache.

[davidmurdoch]

Because we can't checkpoint() before calling runBlock (since: eddd9db#diff-bbc749b6f8c7177c7960ea582daf8f63R371) there are many situations where commit/revert will no longer work (as of v2.5.0).

Clearing the storage cache when it becomes invalid seems like an appropriate fix to me, as the accounts cache is currently cleared when it becomes invalid: 09b8bfb.

[holgerd77]

//cc @mattdean-digicatapult

[mattdean-digicatapult]

This looks like a bug exactly as described by @davidmurdoch. The simplest solution is as described; just clear the storage caches in setStateRoot whenever we clear _cache.