* removing timeout on wait for activation, instead switched to an event driven approach
* fixing unit tests
* linting
* simplifying return
* adding sleep for the remaining slot to avoid cpu spikes
* removing ifstatement on log
* removing ifstatement on log
* improving switch statement
* removing the loop entirely
* fixing unit test
* fixing manu's reported issue with deletion of json file
* missed change around writefile at path
* gofmt
* fixing deepsource issue with reading file
* trying to clean file to avoid deepsource issue
* still getting error trying a different approach
* fixing stream loop
* fixing unit test
* Update validator/keymanager/local/keymanager.go
Co-authored-by: Manu NALEPA <enalepa@offchainlabs.com>
* fixing linting
---------
Co-authored-by: Manu NALEPA <enalepa@offchainlabs.com>
* Initial code for head event streaming
* handle events and error
* keepalive event
* tests
* generate new mock
* remove single case select
* cleanup
* explain eventByteLimit
* use 2 channels in test
* review
* more review
---------
Co-authored-by: james-prysm <90280386+james-prysm@users.noreply.github.com>
* init
- getLocalPayload does not use the proposer ID from the cache but takes
it from the block
- Fixed tests in blockchain package
- Fixed tests in the RPC package
- Fixed spectests
EpochProposers takes 256 bytes that can be avoided to be copied, but
this optimization is not clear to be worth it.
assginmentStatus can be optimized to use the cached version from the
TrackedValidatorsCache
We shouldn't cache the proposer duties when calling getDuties but when
we update the epoch boundary instead
* track validators on prepare proposers
* more rpc tests
* more rpc tests
* initialize grpc caches
* Add back fcu log
Also fix two existing bugs wrong parent hash on pre Capella and wrong
blockhashes on altair
* use beacon default fee recipient if there is none in the vc
* fix validator test
* radek's review
* push always proposer settings even if no flag is specified in the VC
* Only register with the builder if the VC flag is set
Great find by @terencechain
* add regression test
* Radek's review
* change signature of registration builder
* Hook to slot stream instead of block stream on the VC
* Implement StreamSlots in the BN
* mock update
* fix tests
* don't return from stream
* Terence's review
* deepsource second complain
---------
Co-authored-by: rkapka <rkapka@wp.pl>
* `filterAndCacheActiveKeys`: Add test cases
- Validator is in unknown status (to be fitered out)
- Validator is in pending status,
with activation period > current period (to be filtered out)
- Validator is in pending status,
with activation period == current period (to be kept)
* `filterAndCacheActiveKeys`: Keep exiting keys
Initially:
-------
If a validator is in exiting (so, with status==EXITING != ACTIVE) state,
it will be filtered out by the `filterAndCacheActiveKeys` function.
The validator won't be registered to the beacon node.
If this exiting validator has to propose a block:
- the block will be proposed using local block building only.
- the fee recipient will be the one set in the beacon node.
(Additionally, if the beacon node Prysm without any
fee recipient defined at the beacon node level, the fee recipient
will default on the `0x00000...` burn address.)
This commit modifies the `filterAndCacheActiveKeys` function
by stopping filtering out exiting validators.
* making needed changes to beacon API based on removal of blobsidecar from block contents
* fixing tests and reverting some changes to be addressed later
* fixing generated code from protos
* gaz
* fixing get blob handler and adding blob storage to the blob service
* updating unit tests
* WIP
* wip tests
* got tests passing but needs cleanup
* removing gomod and gosum changes
* fixing more tests
* fixing more tests
* fixing more tests
* gaz
* moving some proto types around
* removing unneeded unit test
* fixing proposer paths
* adding more tests
* fixing more tests
* improving more unit tests
* updating one blob only unit test
* changing arguments of buildBlobSidecar
* reverting a change based on feedback
* terence's review items
* fixing test based on new develop changes
* radek's comments
* addressed more comments from radek
* adding in blobs to test data
* fixing casing in test
* removing extra line
* fixing issue from bad merge
* Update beacon-chain/rpc/eth/beacon/handlers_test.go
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* Update beacon-chain/rpc/eth/beacon/handlers_test.go
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* Update beacon-chain/rpc/eth/beacon/handlers_test.go
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* Update beacon-chain/rpc/eth/blob/handlers.go
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* moving core getblob business logic to blocker based on radek's comment
* fixing mock blocker
* gaz
---------
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* `TestStore_GenesisValidatorsRoot_ReadAndWrite`: Make all test cases independents.
In a test with multiple test cases, each test case should be independents.
(aka: Removing test case `A` should not impact test case `B`)
* `SaveGenesisValidatorsRoot`: Allow to overwrite the genesis validator root if the root is the same.
* `ProposalHistoryForSlot`: Add `signingRootExists`
Currently, it is not possible with `ProposalHistoryForSlot` to know if a
proposal is stored with and `0x00000....` signing root or with an empty
signing root. Both cases result to `proposalExists == true` and
`signingRoot == 0x00000`.
This commit adds a new return boolean: `signingRootExists`.
If a proposal has been saved with a `0x00000...` signing root, then:
- `proposalExists` is set to `true`, and
- `signingRootExists` is set to `true`, and
- `signingRoot` is set to `0x00000...`
If a proposal has been saved with an empty signing root, then:
- `proposalExists` is set to `true`, and
- `signingRootExists` is set to `false`, and
- (`signingRoot` is set to `0x00000...`)
* `ImportStandardProtectionJSON`: When importing EIP-3076 Slashing Protection Interchange Format, do not filter any more slashable keys.
Note: Those keys are still saved into the black-listed public keys list.
There is two reason not to do so:
- The EIP-3076 test cases do not know about Prysm's internal black-listed public keys list.
Tests will expect, without looking into this internal black-listed public keys list,
to deny a further signature. If we filter these keys from the DB (even if we keep them
into the black-listed keys list), then some tests will fail.
- If we import a interchange file containing slashable keys and we filter them, then,
if we re-export the DB, those slashing offences won't appear in the exported interchange
file.
* `transformSignedBlocks`: Store an 0-len byte slice
When importing an EIP-3076 interchange format, and when no
signing root is specified into the file, we currently store a
`0x00000.....` signing root.
In such a case, instead storing `0x00000...`, this commit stores
a 0-len byte array, so we can differentiate real `0x000.....` signing
root and no signing-root at all.
* `slashableProposalCheck`: Manage lack of sign root
Currently, `slashableProposalCheck` does not really make a difference
between a `0x0000.....` signing root and a missing signing root.
(Signing roots can be missing when importing an EIP-3076 interchange
file.)
This commit differentiate, for `slashableProposalCheck`, `0x0000....`
signing root and a missing signing root.
* `AttestationRecord.SigningRoot`: ==> `[]byte`
When importing attestations from EIP-3076 interchange format,
the signing root of an attestation may be missing.
Currently, Prysm consider any missing attestation signing root as
`0x000...`.
However, it may conflict with signing root which really are equal to
`0x000...`.
This commit transforms `AttestationRecord.SigningRoot` from `[32]byte` to
`[]byte`, and change the minimal set of functions (sic) to support this
new type.
* `CheckSlashableAttestation`: Empty signing root
Regarding slashing roots, 2 attestations are slashable, if:
- both signing roots are defined and differs, or
- one attestation exists, but without a signing root
* `filterSlashablePubKeysFromAttestations`: Err sort
Rergarding `CheckSlashableAttestation`, we consider that:
- If slashable == NotSlashable and err != nil, then CheckSlashableAttestation
failed.
- If slashable != NotSlashable, then err contains the reason why the attestation
is slashable.
* `setupEIP3076SpecTests`: Update to `v5.3.0`
This commit:
- Updates the version of EIP-3076 tests to `v.5.2.1`.
- Setups on anti-slashing DB per test case, instead per step.
* `ImportStandardProtectionJSON`: Reduce cycl cmplxt
* `AttestationHistoryForPubKey`: copy signing root
BoltDB documentation specifies:
| Byte slices returned from Bolt are only valid during a transaction.
| Once the transaction has been committed or rolled back then the memory
| they point to can be reused by a new page or can be unmapped
| from virtual memory and you'll see an unexpected fault address panic
| when accessing it.
* WIP
* fixing tests
* fixing bazel
* fixing api client
* fixing tests
* fixing more tests and bazel
* fixing trace and more bazel issues
* fixing router path function definitions
* fixing more tests and deep source issues
* adding delete test
* if a route is provided, reregister before the catch all on the middleware.
* fixing linting
* fixing deepsource complaint
* gaz
* more deepsource issues
* fixing missed err check
* changing how routes are registered
* radek reviews
* Update validator/rpc/handlers_keymanager.go
Co-authored-by: Sammy Rosso <15244892+saolyn@users.noreply.github.com>
* Update validator/rpc/handlers_keymanager.go
Co-authored-by: Sammy Rosso <15244892+saolyn@users.noreply.github.com>
* fixing unit test after sammy's review
* adding radek's comments
---------
Co-authored-by: Sammy Rosso <15244892+saolyn@users.noreply.github.com>
* Fix block proposals in the REST validator client
* fix graffiti test
* return empty graffiti
* fallback to old endpoints
* logs
* handle 404
* everything passes
* review from James
* log undecoded value
* test fixes and additions
---------
Co-authored-by: prylabs-bulldozer[bot] <58059840+prylabs-bulldozer[bot]@users.noreply.github.com>
This cancel function is currently a no-op due to the blank identifier.
One might argue that the cancel func should be restored from no-op by
replacing the blank identifier with the proper variable. When the parent
context is cancelled, however, all the functions down the call tree with
the context will be notified of the cancellation anyway. Removing the
cancel function would not change any outcome under the current
implementation.
Co-authored-by: Preston Van Loon <pvanloon@offchainlabs.com>
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* Fix exported names that start with a package name
* A few more renames
* Fix exported names that start with a package name
* A few more renames
* Radek's feedback
* Fix conflict
* fix keymanager test
* Fix comments
---------
Co-authored-by: prylabs-bulldozer[bot] <58059840+prylabs-bulldozer[bot]@users.noreply.github.com>
