blockchain: Add invalidate/reconsider infrastructure.

[davecgh]

This requires #2518.

This adds infrastructure to support invalidating arbitrary blocks as if they had violated a consensus rule as well as reconsidering arbitrary blocks for validation under the current consensus rules and includes comprehensive tests to ensure proper functionality.

It also adds two new rpc commands invalidateblock and reconsiderblock to expose the new functionality and updates the JSON-RPC API documentation accordingly.

Example use cases this enables:

• Revalidating blocks under new consensus rules
  • Consider the case of old software rejecting blocks due to new consensus rules activating on the network and then upgrading to a new version that supports the new rules
• Possibility of manually generating snapshots of historical state such as live tickets and available utxos
• Manually recovering from unexpected circumstances
• Easier chain reorganization testing

---

Testing notes:

In order to make the new commands available to dcrctl, a replace directive to use the updated and unreleased types module is needed. The following command from the dcrctl directory will accomplish that:

go mod edit -replace=github.com/decred/dcrd/rpc/jsonrpc/types/v2=../dcrd/rpc/jsonrpc/types

[davecgh]

Here is an example of the PR in action on simnet:

Invalidate and reconsider same block:

$ for i in $(seq 28 1 32); do echo "$(./ctl getblockhash $i) (height $i)"; done
19630e8a7bd723f0d37e7f76b824f43e878ae005063007248220350365bf42d9 (height 28)
0e48bb52098b5bc014f94e42543504f4f5f799aa96601a8ba5e8717632a1ba06 (height 29)
04e9a016058439f4f19c16f9b316dd52259258baf9e0067a311b0ed33c739569 (height 30)
0692d68e9abb2c9049a10e98fa11789df21b90bbd4fe3191d833826117b2040e (height 31)
00c1e024cdadb091e3f24b40c4a73f51704a60b8fbc4a735725c3c1577a20f7a (height 32)

$ ./ctl invalidateblock 04e9a016058439f4f19c16f9b316dd52259258baf9e0067a311b0ed33c739569  # height 30
$ ./ctl getbestblockhash | ./ctl getblockheader - | jq '{height,hash}'
{
  "height": 29,
  "hash": "0e48bb52098b5bc014f94e42543504f4f5f799aa96601a8ba5e8717632a1ba06"
}
$ ./ctl reconsiderblock 04e9a016058439f4f19c16f9b316dd52259258baf9e0067a311b0ed33c739569  # height 30
$ ./ctl getbestblockhash | ./ctl getblockheader - | jq '{height,hash}'
{
  "height": 32,
  "hash": "00c1e024cdadb091e3f24b40c4a73f51704a60b8fbc4a735725c3c1577a20f7a"
}

Nested invalidate and reconsider:

$ ./ctl invalidateblock 0692d68e9abb2c9049a10e98fa11789df21b90bbd4fe3191d833826117b2040e  # height 31
$ ./ctl invalidateblock 0e48bb52098b5bc014f94e42543504f4f5f799aa96601a8ba5e8717632a1ba06  # height 29
$ ./ctl getbestblockhash | ./ctl getblockheader - | jq '{height,hash}'
{
  "height": 28,
  "hash": "19630e8a7bd723f0d37e7f76b824f43e878ae005063007248220350365bf42d9"
}
$ ./ctl reconsiderblock 0e48bb52098b5bc014f94e42543504f4f5f799aa96601a8ba5e8717632a1ba06  # height 29
$ ./ctl getbestblockhash | ./ctl getblockheader - | jq '{height,hash}'
{
  "height": 30,
  "hash": "04e9a016058439f4f19c16f9b316dd52259258baf9e0067a311b0ed33c739569"
}
$ ./ctl reconsiderblock 0692d68e9abb2c9049a10e98fa11789df21b90bbd4fe3191d833826117b2040e  # height 31
$ ./ctl getbestblockhash | ./ctl getblockheader - | jq '{height,hash}'
{
  "height": 32,
  "hash": "00c1e024cdadb091e3f24b40c4a73f51704a60b8fbc4a735725c3c1577a20f7a"
}

Reconsider descendant of one or more invalidated ancestors:

$ ./ctl invalidateblock 0692d68e9abb2c9049a10e98fa11789df21b90bbd4fe3191d833826117b2040e  # height 31
$ ./ctl getbestblockhash | ./ctl getblockheader - | jq '{height,hash}'
{
  "height": 30,
  "hash": "04e9a016058439f4f19c16f9b316dd52259258baf9e0067a311b0ed33c739569"
}
$ ./ctl invalidateblock 0e48bb52098b5bc014f94e42543504f4f5f799aa96601a8ba5e8717632a1ba06  # height 29
{
  "height": 28,
  "hash": "19630e8a7bd723f0d37e7f76b824f43e878ae005063007248220350365bf42d9"
}
$ ./ctl reconsiderblock 00c1e024cdadb091e3f24b40c4a73f51704a60b8fbc4a735725c3c1577a20f7a  # height 32
$ ./ctl getbestblockhash | ./ctl getblockheader - | jq '{height,hash}'
{
  "height": 32,
  "hash": "00c1e024cdadb091e3f24b40c4a73f51704a60b8fbc4a735725c3c1577a20f7a"
}

[rstaudt2]

This is awesome. I had a few comments inline, and outside of that, everything looks good. I used this to test deep reorgs (up to thousands of blocks deep) against real mainnet data, which wasn't really possible before. I also tested that scenario against #2540 and was able to reproduce an issue, so this has already proven to be very useful!

[dnldd]

Tested on simnet, working as expected. These RPCs are really nice! 👍

[davecgh]

I added a couple of logging statements so the commands are not silent and rebased.

[rstaudt2]

Looks good. Retested with the latest, and everything worked as expected, including the newly added log.

[JoeGruffins]

When reconsidering, I get a warning for every block: [WRN] FEES: Trying to process mined transactions at block 594631 when previous best block was at height 594641

[JoeGruffins]

If I manually invalidate a block, is it possible to undo that without remembering the block hash? If not an rpc call that shows what I have manually disabled would be handy.

[sefbkn]

Ran this on mainnet without issue. When started back up, it syncs back to the best height on the network when dcrd is stopped while invalidating/reconsidering arbitrary blocks.

Just have a couple of minor comments.

[davecgh]

When reconsidering, I get a warning for every block: [WRN] FEES: Trying to process mined transactions at block 594631 when previous best block was at height 594641

@JoeGruffins Yes, that is normal. The fees code doesn't update on reorgs. It probably shouldn't be a warning, but that is not related to this PR, so I'll leave it for another.

[davecgh]

If I manually invalidate a block, is it possible to undo that without remembering the block hash?

Yes. You can just use getchaintips to find the latest tip hash and then reconsider that tip. This intentionally reconsiders all ancestors too which is largely in part to allow exactly this.

If not an rpc call that shows what I have manually disabled would be handy.

Given the above, I don't think this is really necessary. However, an RPC to show information regarding invalid blocks is not a bad idea! That said, it would have to come with some caveats. Namely, you can't distinguish between a block that was manually marked invalid and one that actually failed consensus validation and this is by design. So, such an RPC would really need to show all blocks that have failed validation which would include both types.

It's not something I would want to do as a part of this PR, so it would be something for another day.