Issue before this commit:
The `currentSlot` in the `getSubnetsToSubscribe` is really set to the current slot. So, even if the `registerSubscribers` function is called
one epoch in advance, as done in `registerForUpcomingFork`, then
`currentSlot` will still be - really - the current slot.
==> The new blobs subnets will be subscribed only at the start of the
Electra fork, and not one epoch in advance.
With this commit:
The new blobs subnets will be effectively subscribed one epoch in advance.
* Re-organize thet content of the `*.proto` files.
The content of the `*.proto` files is sorted by hard fork,
then with a top-down fashion.
Sorting first by hard fork lets the reader to easily see new or modified fields.
Then, sorting with a top-down fashion lets the user to first see the big picture,
then to dive into details.
Also, the `new in <hard fork>` mentions are only written for the given hard fork.
Thus, it'll avoid in the future the majority of the fields, not initially
present in phase 0, to have the `new in <hard fork> mention`.
This commit does not bring any new functional change.
* Update proto/prysm/v1alpha1/attestation.proto
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* Update proto/prysm/v1alpha1/beacon_state.proto
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* Fix Radek's comment.
---------
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* Use 16 bit random value
* enforce 0x02 credentials for consolidations
* Limit consolidating balance by validator effective balance
* Update max blob commitment size
* Fix next sync committee indices
* Bytes to little endian
* Handle proposer computations in between forks
* Fix config and tests
Fix tests
Fix tests
* Fix test stream events by properly set effective balance
Fix test stream events by properly set effective balance
* Preallocate buffers to avoid repeated allocations
* Potuz's feedback
* Use 16 bit random value
* enforce 0x02 credentials for consolidations
* Limit consolidating balance by validator effective balance
* Update max blob commitment size
* Fix next sync committee indices
* Bytes to little endian
* Handle proposer computations in between forks
* Fix config and tests
Fix tests
Fix tests
* Fix test stream events by properly set effective balance
Fix test stream events by properly set effective balance
* Preallocate buffers to avoid repeated allocations
* Potuz's feedback
* Fix change log
* Rename existing AttestationCache
* Cache with simple Add
* fix import cycle
* logic for unaggregated, aggregated and block attestations
* some small fixes
* remove Seen
* finishing touches
* feature flag
* extract forkchoice atts to separate type
* gate new functionality behind feature flag
* revert test files
* preparing for review
* change Id to [32]byte
* Potuz's review
* Potuz's review pt 2
* Nishant's review
* keep flat list of atts
* fix ForkchoiceAttestations() function
* Tests for Add, GetAll, Count
* Tests for remaining functions
* use DeepEqual
* fix tests
* documentation
* changelog <3
* v2 handlers
* nil check for forkchoice atts
* guard against 0 bits set
* fix failing test
* Preston's review
* better godocs
* move flag settings out of individual tests
* add tests with better or worst old update
* remove whitespace
* add zero fields for default update at deneb block in electra
---------
Co-authored-by: Radosław Kapka <rkapka@wp.pl>
* Refactor subnets subscriptions.
* Remove totally static/dynamic distinction.
* Unsubscribing from topic: Use INFO instead of log.
==> So we have something symmetrical with subscriptions.
* Address Nishant's comment.
Rationale:
Before this commit, the internal loop exited if:
- the expected amount of peers is found, or,
- the iterator returns `false` (exhaustion), or
- `batchSize` iterations are done.
The issue with the iterations count is, in case not enough
peer are found AND `iterator.Next` always returns `true`,
we don't control WHEN the loop is going to stop.
The root cause is we don't control the time needed to
run the `iterator.Next` function, which is a function of
`devp2P (geth)`.
The value of `batchSize (2000)` was chosen arbitrarily.
It turns out the time needed to run `iterator.Next` can go from a few micro seconds to a few hundreds of milliseconds.
==> In small networks (example: E2E tests), it was possible for the loop not to exit during several dozen of seconds.
With this commit, we replace the `batchSize` by a `batchPeriod`, ensuring the loop will never
run longer than `batchPeriod`, even in a small network.
Co-authored-by: Nishant Das <nishdas93@gmail.com>
* Add missed exit checks to consolidation processing
* Use safe add
* gaz
* Pass spec tests (except single attestation)
Revert params.SetupTestConfigCleanupWithLock(t)
* Update earlist exit epoch for upgrade to electra
* Validate that each committee bitfield in an aggregate contains at least one non-zero bit
* Add single attestation
* Add single attestation to ssz static
* Fix typo
Co-authored-by: Md Amaan <114795592+Redidacove@users.noreply.github.com>
* Update UpgradeToElectra comments
* Add no lint dupword
---------
Co-authored-by: james-prysm <james@prysmaticlabs.com>
Co-authored-by: Md Amaan <114795592+Redidacove@users.noreply.github.com>
* Make it the default
* Changelog
* Remove outdated flag
* Update `go-libp2p` to `v0.36.5` and `webtransport-go` to `master`.
---------
Co-authored-by: Manu NALEPA <enalepa@offchainlabs.com>
* `listenForNewNodes` and `FindPeersWithSubnet`: Stop using `Readnodes` and use iterator instead.
It avoids infinite loop in small devnets.
* Update beacon-chain/p2p/discovery.go
Co-authored-by: Sammy Rosso <15244892+saolyn@users.noreply.github.com>
---------
Co-authored-by: Sammy Rosso <15244892+saolyn@users.noreply.github.com>
