Metadata:

- Fixes ENG-1662
- Fixes ENG-1664
- Fixes ENG-1663

Adds the actor testing framework and implements some basic actor tests. The tests are:

- An actor that makes repeated deposits
- An actor that makes repeated transfers

For information on how the framework works, see [integration-tests/actor-tests/README.md](integration-tests/actor-tests/README.md).

for reference these were moved to https://github.com/ethereum-optimism/optimistic-specs/tree/move-legacy-monorepo-specs

fix: don't round down average gas per epoch

Version Packages

Develop -> Master PR

fix: trigger node 16 releases

Upgrade the L1 hardhat node docker
image to use node version 16

changesets: don't track builder image

Tracking the builder image causes problems
in the CI. It started to be tracked when
there was a desire for faster builds. Publishing
a versioned builder image has resulted in
many CI publishing related bugs. This commit
removes tracking of the builder image.

Develop -> Master Release PR

Version Packages

Develop -> Master PR

docker: update node version to 16

Develop -> Master Release PR

Version Packages

Develop -> Master PR

Also corrects the targetGps constant in the test harness. The constant
was chosen to be the expected Gps of the initial burst, causing the Gps
to stablize in the second part of the test at the floor.

Adding a check that the average measured gas price matches our desired
constant prevents this from silently passing.

feat: add batch-tx submitter custom encoding and test vectors

fix: minor improvements to empty sequencer block handling

Prior to this commit, an empty block would be submitted to the taskCh
before sanity check whether or not it was empty. This assertion is moved
up to fail before submitting the task.

It seems the placement of the check before was done to get around the
StreamUncle unit test, which would allow commit to submit an empty task
before ultimately failing. Now that the test is modified to reflect the
expected behavior of the sequencer, this placement is no longer
necessary.

This commit fixes a bug in the way txCount was used for detecting and
aborting empty blocks. Previously, txCount was incremented immediately
after popping off the new transaction but before it is executed. This
was incorrect, as the transaction can still be rejected for nonce/gas
errors later on. Instead, we modify the logic only count transactions
that are successfully added, and utilize the existing w.current.tcount
member to implement the detection.

With empty block mining logic totally removed in the prior commit,
worker.commit is now only called with update=true. We can now safely
remove the argument and the affected codepaths.

As of the prior commit, all calls to the `commit` pass a noempty value
of true, indicating that the sequencer should not start mining empty
blocks. Therefore, we can remove all usage of noempty from the worker.

When using PoW, geth begins mining on empty blocks before attempting to
select which transactions to include. This optimization is not needed,
as we require that every L2 block have exactly one transaction. This
commit disables this behavior an modifies the tests accordingly. The
test is also expanded to be stricter wrt to testing number of txs and
uncles on all blocks.

Slither github action fix and ignore database