op-supervisor,op-node: Handle non-genesis Interop

[sebastianst]

Description

The op-node and op-supervisor are taught to handle a non-Genesis Interop activation timestamp properly in managed mode. In essence, the op-supervisor just does less while the op-node does a bit more pre-Interop:

• Before Interop is active, the op-node locally auto-promotes local-(un)safe to cross(un)safe blocks (as before).
• Only L1 derivation is driven by the op-supervisor pre-Interop.
  • ExhaustL1 events are answered by new L1 blocks as they become available.
  • Reset requests are answered by the supervisor with a special ResetPreInterop response.
• When the supervisor detects the (unsafe and safe) activation blocks, it initiates the (events and derivation) DBs.
• If a post-Interop reset bisection cannot find a consistent block, i.e. not even the anchor block stored in the DB is consistent with the node's view, then a pre-Interop reset is initiated as well.

The explicit notion of the anchor block is also mostly removed already in this PR.

• The full supervisor config set includes genesis information, so the op-supervisor tests at startup whether the genesis block already has Interop active. If so, it will locally initiate the DB from genesis information, not querying AnchorBlock on the managed nodes.
• If genesis hasn't Interop active yet, it is assumed that the activation block passes by the supervisor via its block subscription on its manages nodes.

This PR also makes the logs DB check itself whether it is initialized so that the ChainsDB initialized markers are only used to denote a fully initialized ChainsDB from a safe activation block.

Tests

The action test TestInteropUpgrade at op-e2e/actions/interop/interop_fork_test.go already runs Interop through an activation and asserts that the Interop activation block is properly sequenced and derived. This implicitly confirms that the activation logic in op-node/op-supervisor functions properly around the Interop activation timestamp. Some modifications were done to this test since it had some false logic that only worked with the previous pseudo-pre-Interop logic that the supervisor was following.

A post-genesis Interop variant of TestBackendLifetime was also added, and the previous tests adapted.

More action and acceptance tests will be added in future work.

Additional context

In a followup, full reorg support around the activation block is added. It probably wouldn't work right now, will be worked out in #15774

A few proofs tests or checks had to be commented out because the op-challenger (and probably other bits) still need some more work to make non-genesis activation work. This will be resolved with #16166

This PR breaks mutli-node support, which we decided to drop. So a few tests had to be skipped because of this. They will either be removed or fixed (#16187).

This PR also doesn't fully remove the notion of the anchor block yet. This is reserved for future work #16140

Metadata

Fixes #13732

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 96.24%. Comparing base (e7e42d2) to head (b968b00).
Report is 3 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff              @@
##           develop   #16008       +/-   ##
============================================
+ Coverage    81.54%   96.24%   +14.69%     
============================================
  Files          161      106       -55     
  Lines         8858     4575     -4283     
============================================
- Hits          7223     4403     -2820     
+ Misses        1502      172     -1330     
+ Partials       133        0      -133     

Flag	Coverage Δ
cannon-go-tests-64	?
contracts-bedrock-tests	96.24% <ø> (ø)

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

see 55 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[ajsutton]

Took a look at the challenger failures and it looks like this change has broken the key requirement challenger has on it's API that optimism_superRootAtTimestamp will return a valid super root for any timestamp up to at least the current safe head. It is critical that it can return a super root even prior to interop activating, because the new anchor state we set will be from before interop hard fork activates and any block after that can be disputed. So challenger has to be able to get the valid super root for timestamps from that anchor state forwards in order to post correct claims.