Marge Bot (5aeb1b5f) at 20 Mar 15:26
Merge !21123: Etherlink/Benchmark: workaround stale stor...
... and 1 more commit
Marge Bot (0e9adbcb) at 20 Mar 15:26
The Etherlink tezt cloud benchmark crashes when using a custom kernel with a mainnet snapshot. The EVM node exits with "No value found under /evm/sequencer" shortly after producing the first block.
When a custom kernel triggers a storage migration (e.g. V50, which relocates the sequencer key to a new path), the EVM node's in-memory cached storage version becomes stale. The node keeps looking for the sequencer key at the old path, which the kernel has already deleted. This only happens when the kernel is replaced at init time rather than through the normal upgrade mechanism, because the storage version refresh is only wired to scheduled upgrades.
Relates to L2-986.
After the first block is produced (which triggers the kernel migration), the benchmark now restarts the EVM node. On restart, the node re-reads the storage version from the current state and picks up the correct paths. This is a temporary workaround until the node itself is patched.
Marge Bot (05340a60) at 20 Mar 11:25
This is an instance of the typo train, an MR grouping typo fixes and other minor (and not so urgent) changes to the doc.
See meta-issue #2329 for more details.
Encourage doc fixes from anyone.
Nothing is specified in advance! Rather, doc fixes are welcome from anyone, as long as in Draft mode.
Read the changes and check that they are correct and they bring some improvements.
Reviewers field below.Assignee the next person who should take action on that MR
Marge Bot (b0a04807) at 20 Mar 11:23
Marge Bot (b78df87d) at 20 Mar 11:23
Merge !21297: TezosX: Fix tezos operation injection
... and 1 more commit
Tezos operation injection was broken in tezt scenarios as demonstrated by !21296.
This fail is caused by 2 things:
Instant confirmations are enabled by default for tezt scenarios when:
~preconfirmation_stream_enabled: (Option.value ~default:(not enable_multichain) instant_confirmations)
And tick() function of the tx queue (used to push received transaction to the block producer), is not called for tezos operation injection.
Call already present tick() function in etherlink/bin_node/lib_dev/rpc_server.ml after add_operation.
Test writter in !21296 by @spalmer25 passes after being rebased on this MR.
I can't assign this MR to a merge train!
Here are the error(s) I have detected:
Note that there are some errors I can't detect. So please have a look at the merge checks done by Gitlab. (*)
If you need help, please contact the infrastructure team on Slack.
(*) Near top of the MR page, between approvals and activity.
Marge Bot (9155191d) at 20 Mar 10:32
Marge Bot (1de15acd) at 20 Mar 10:32
Merge !20591: Build: Check opam version before editing a...
... and 1 more commit
Marge Bot (fe0a8b64) at 20 Mar 10:32
Checks that opam has a minimum version before changing the archive-mirrors config option
opam 2.1 will fail when trying to set this option
docs/protocols/alpha.rst for the protocol and the environment, CHANGES.rst at the root of the repository for everything else).Reviewers field below.Assignee the next person who should take action on that MR
Marge Bot (b77812c2) at 20 Mar 10:32
Merge !21201: Rollup node: add --level option to snapsho...
... and 7 more commits
Add a --level option to the smart rollup node's snapshot export and import commands.
For export, it controls the last level included in the snapshot. It accepts head, finalized, cemented, or a numeric level, and defaults to finalized.
For import, it accepts a numeric level and allows importing a snapshot at an earlier level than the one provided by the snapshot.
Snapshot imports can fail when the L2 head stored in the snapshot sits on a branch that has since been reorganized on L1. The commitment at that level (if any) may never be published, causing the post-import validation to reject the snapshot.
Defaulting export to finalized avoids producing snapshots with heads that are likely to be reorganized. The import option provides a recovery path when a snapshot's head is already on an orphaned branch.
A new snapshot_level CLI type is added with the accepted shortcuts and numeric values. On export, the store is truncated to the requested level after the vacuum copy. On import, the store is reset to the requested level after extraction, and the commitment publication check is adjusted to account for the new head.
Marge Bot (f139f280) at 20 Mar 09:57
Merge !21267: TezosX: Share access list across CRACs
... and 2 more commits
Marge Bot (3dec08eb) at 20 Mar 09:57
Keep the EIP-2930 access list through cross runtime calls.
In the EVM > Michelson > EVM scenario, origin access list is currenlty lost. This is not something that is handled naturally by sharing the REVM journal. We must handle it explicitely. Closes https://linear.app/tezos/issue/L2-983/handle-access-list-in-evm-runtime.
TransactionOrigin::UserInput now contains an access_list. We chose not to make the update available to the CRAC user because it is:
EvmJournal stores the access_list
UserInput case of CRACTxEnvBuilder which also requires ownershipMarge Bot (0604453a) at 20 Mar 09:45