copydb panic because it can't retrieve the state tree in current height - 64 block

[huangyg11]

The behavior of copydb ( or fast sync) is to sync state tree in the block of current height - 64, but it seems geth don't persist state tree in the block of current height - 64 before exist (it only persist current height, current height - 1, current height - 127) . So copydb will panic.

If I panic in copydb, I have to start geth and wait it syncing another 64 blocks and exist. Then invoke copydb again, it will work normally.

[fjl]

This needs a fix in eth/downloader because copydb uses it to "fast sync" into the local database.

[holiman]

I've trie to repro this, switching the pivot point back and forth. Everything seems to work, although I agree it should fail. However, I don't see why it would panic, so it would be good to see the stack trace of such a panic.

(yeah I know it's an old issue, sorry for not picking it up sooner)

[karalabe]

The command was dropped on master. We have (manual) state pruning shipped, which is better and faster.