This changelog is a work in progress and may contain notes for versions which have not actually been released. Check the Releases page to see full release notes and more information about the latest released versions.
- Improved the performance of validating order filters in the browser #809.
- Mesh now ensures on startup that the chain ID of your Ethereum RPC endpoint matches config.EthereumChainID #733.
- Fixed a compatibility issue in
@0x/mesh-browser-lite
for Safari and some other browsers #770. - Fixes an issue that would allow expired orders to be returned in
GetOrders
773 - Fixed a rare bug where ERC721 approval events could be missed #782
- Fixed a critical bug in the ordersync protocol which resulted in only 50% of existing orders being shared when a new peer joins the network. New orders are shared separately and were unaffected. #760.
- Greatly reduced latency for propagating orders, especially for browser nodes #756.
- Added support for
checkGasPrice
StaticCall asset data #744
- Improved speed and efficiency of peer discovery, especially when using custom order filters #729.
- Added a lightweight package to use for loading Mesh's Wasm binary in a streaming manner #707.
- Fixed an issue where incoming orders could sometimes be dropped by peers #732.
- Fix bug where we weren't enforcing that we never store more than
miniHeaderRetentionLimit
block headers in the DB. This caused issue #667 and also caused the Mesh node's DB storage to continuously grow over time. (#716)
- As a result of implementing custom order filters, some of the code Mesh uses under the hood to share orders with peers has changed. As a result this version of Mesh cannot share orders with any older versions and vice versa (#630).
- Implemented a new protocol for sharing existing orders with peers. This will drastically reduce bandwidth and CPU usage and increase the speed at which new orders are propagated. (#692).
- Rename
RPC_ADDR
toWS_RPC_ADDR
since we now support both WS and HTTP JSON-RPC endpoints. (#658)
- Implemented custom order filters, which allow users to filter out all but the orders they care about. When a custom order filter is specified, Mesh will only send and receive orders that pass the filter. (#630).
- Developers can now override the contract addresses for any testnet using the
CUSTOM_CONTRACT_ADDRESSES
env config (#640). - Added
getOrdersForPageAsync
method to@0x/mesh-rpc-client
WS client interface so that clients can paginate through the retrieved orders themselves (#642). - Added support for passing in your own Web3 provider when using the
@0x/mesh-browser
package. (#665). - Add support for orders involving Chai ERC20Bridge assetData (#663)
- Add support for calling JSON-RPC methods over HTTP (env config
HTTP_RPC_ADDR
defaults tolocalhost:60556
). (#658)
- Fixed some of the browser typescript bindings to be consistent with the Go and smart contract implementations (#697).
- Added
getStatsAsync
to the@0x/mesh-browser
package. (#654). - Added
getOrdersAsync
andgetOrdersForPageAsync
to the@0x/mesh-browser
package. (#655).
- Update DevUtils contract address to fix intermittent revert issues. (#671).
- Update DevUtils contract to version that removed maker transfer simulation. (#662).
- Fix faulty Go to Javascript conversion logic. (#659).
- Updated dockerfiles to work with Go modules. (#646).
- Update DevUtils mainnet contract address to version that fixes MAP order validation issue (#644).
- Fixed a regression which can result in memory leaks. (#650).
- Reduced startup time for Mesh node by only waiting for a block to be processed if Mesh isn't already sync'ed up to the latest block. (#622)
- Increased the maximum size for encoded orders from ~8kB to 16kB (#631).
- Fixed a typo ("rendervouz" --> "rendezvous") in GetStatsResponse. (#611).
- Fixed a bug where we attempted to update the same order multiple times in a single DB txn, causing the later update to noop. (#623).
- Fixed a bug which could cause Mesh to exit if a re-org condition occurs causing a block to be added and removed within the same block sync operation. (#614).
- Changed the response from
@0x/mesh-ts-client
'sgetOrdersAsync
endpoint to include thesnapshotID
andsnapshotTimestamp
at which the Mesh DB was queried along with the orders found. (#591) - Increased the default
ETHEREUM_RPC_MAX_REQUESTS_PER_24_HR_UTC
from 100k to 200k (#596).
- Instead of progressing Mesh forward by a single block on every invocation of the
BLOCK_POLLING_INTERVAL
, we now attempt to sync as many blocks as necessary to reach the latest block available. This will reduce the chances of Mesh becoming out-of-sync with it's backing Ethereum node (#564) - Added a new environment variable
ENABLE_ETHEREUM_RPC_RATE_LIMITING
and config optionenableEthereumRPCRateLimiting
which can be used to completely disable Mesh's internal Ethereum RPC rate limiting features. By default it is enabled, and disabling can have some consequences depending on your RPC provider. (#584) - Added a
SnapshotTimestamp
field toGetOrdersResponse
, the return type of themesh_getOrders
RPC method. This way, the caller can know at what point in time the snapshot had been created. (#591) - Improved batching of events emitted from order events subscriptions (#566)
- Added timestamp to order events (#602)
- Fixed an issue where order updates could have been missed if Mesh discovered blocks but didn't have time to process them before getting shut down. Now, blocks are only persisted to the DB once any order updates resulting from it have been processed. (#566).
- Fixed a race-condition when adding new orders to Mesh which could result in order-relevant events being missed if they occured very soon after the order was submitted and the order validation RPC call took a long time (#566).
- Upgraded the
web3-provider
dependency used by@0x/mesh-rpc-client
in order to fix a bug where it was requiring eitherprocess
ORwindow
to exist in the global scope (#601). - Fixed an issue where the internal Ethereum RPC rate limiter could be too aggressive in certain scenarios (#596).
- Add a default RPC request timeout of 30sec to all non-subscription requests sent by
@0x/mesh-rpc-client
to avoid the client from hanging endlessly if it misses a response due to network disruption (#603).
- Fixed a bug where
makerAssetFeeData
ortakerAssetFeeData
were sometimes being set to an empty string instead of0x
in the browser (#579).
- Fixed a bug which could cause Mesh to crash with a nil pointer exception if RPC requests are sent too quickly during/immediately after start up (#560).
- Update V3 contract addresses for Ganache snapshot. (#569)
- Fixed a bug where the internal order event feed could be come blocked, rendering Mesh unable to receive any new orders or update existing ones (#552).
- Update V3 contract addresses for mainnet/testnets. (#547)
- Added support for persistence in the browser. Users of the
@0x/mesh-browser
package will now be able to retain orders and other relevant parts of the state when refreshing the page or closing and re-opening the browser. (#533).
- Fix bug where Mesh nodes were logging receipt and re-sharing with peers duplicate orders already stored in it's DB, if the duplicate order was submitted via JSON-RPC. (#529)
- Add missing
UNEXPIRED
OrderEventEndState
enum value to both@0x/mesh-rpc-client
and@0x/mesh-browser
and missingSTOPPED_WATCHING
value from@0x/mesh-rpc-client
. - Fixed a potential memory leak by using the latest version of
github.com/libp2p/go-libp2p-kad-dht
(#539). - Changed the default port for
RPC_ADDR
from a random available port to60557
. Some documentation already assumed60557
was the default port. Now all documentation has been updated for consistency with this change. (#542). - Fixed a potential nil pointer exception in log hooks (#543).
- Fixed a bug where successful closes of an rpc subscription were being reported as errors (#544).
- We now log the error and stack trace if an RPC method panics. Before, these errors were swallowed by the panic recovery logic in
go-ethereum
'srpc
package. (#545) - Previously, we used to fast-sync block events missed since a Mesh node was last online. If this was more than 128 blocks ago, the fast-sync would fail if Mesh was not connected to an Ethereum node with the
--archive
flag enabled. We now fast-sync only if less than 128 blocks have elapsed. Otherwise, we simply re-validate all orders and continue processing block events from the latest block. (#407)
- Upgraded
@0x
deps in@0x/mesh-rpc-client
- Upgrade BigNumber dep used by
@0x/mesh-rpc-client
to~9.0.0
(#527)
- Upgrade Mesh's
libp2p/go-flow-metrics
dep in an attempt to fix a bug in the flow metrics module. (#521) - Updates
DevUtils.sol
addresses on all networks to the latest version which fixed a deploy issue. (#520)
- Fixed an oversight which granted immunity from bandwidth banning for any peer using a relayed connection (#509).
- Fixed a typo in the
@0x/mesh-browser
package that resulted in some config options not being passed through correctly (#502). - Fixed a bug in ETH JSON-RPC rate limiter where not all dates were being properly converted to UTC, causing Mesh to malfunction if the local time was a day earlier or later than UTC. (#505)
- Fixed a bug in the TypeScript RPC client that prevented orders from being added (#514).
Note: This release will require wiping your Mesh's DB before upgrading. The DB location defaults to ./0x_mesh/db
.
- Removed
RPC_PORT
environment variable. The newRPC_ADDR
environment variable allows specifying both the interface and port (#487). - Due to security concerns and new rate limiting mechanisms, the default bind address for the RPC API has changed from
0.0.0.0
tolocalhost
. Users who previously did not setRPC_PORT
may need to now manually setRPC_ADDR
to enable other applications to access the RPC API. If you are using Docker Compose, we recommend using links. If you do need to setRPC_ADDR
to bind on0.0.0.0
, please be aware of the security implications and consider protecting access to Mesh via a third-party firewall. (See #444 and #487 for more details). - Changed the
EXPIRED
event such that it is emitted when an order is expired according to the latest block timestamp, not anymore based on UTC time. (#490) - Removed the
EXPIRATION_BUFFER_SECONDS
env config since we no longer compute order expiration using UTC time. (#490)
- Added an
UNEXPIRED
order event kind which is emitted for orders that were previously considered expired but due to a block-reorg causing the latest block timestamp to be earlier than the previous latest block timestamp, are no longer expired. (#490) - Added support for decoding Axie Infinity
Transfer
andApprove
ERC721 events which differ from the ERC721 standard. (#494)
- Fixed a bug in the Go RPC client which resulted in errors when receving order events with at least one contract event (#496).
- Renamed env config from
ETHEREUM_NETWORK_ID
toETHEREUM_CHAIN_ID
sincenetwork
is a misnomer here and what we actually care about is thechainID
. Most chains have the same id for their p2p network and chain. From the ones we support, the only outlier is Ganache, for which you will now need to supply1337
instead of50
(Learn more: https://medium.com/@pedrouid/chainid-vs-networkid-how-do-they-differ-on-ethereum-eec2ed41635b) (#485) - Rejected order code
OrderForIncorrectNetwork
has been changed toOrderForIncorrectChain
(#485)
- Implemented a new strategy for limiting the amount of database storage used by Mesh and removing orders when the database is full. This strategy involves a dynamically adjusting maximum expiration time. When the database is full, Mesh will enforce a maximum expiration time for all incoming orders and remove any existing orders with an expiration time too far in the future. If conditions change and there is enough space in the database again, the max expiration time will slowly increase. This is a short term solution which solves the immediate issue of finite storage capacities and does a decent job of protecting against spam. We expect to improve and possibly replace it in the future. See #450 for more details.
- Added support for a new feature called "order pinning" (#474). Pinned orders will not be affected by any DDoS prevention or incentive mechanisms (including the new dynamic max expiration time feature) and will always stay in storage until they are no longer fillable. By default, all orders which are submitted via either the JSON-RPC API or the
addOrdersAsync
function in the TypeScript bindings will be pinned. - Re-enabled bandwidth-based peer banning with a workaround to deal with erroneous spikes #478.
- Improved the aggressiveness at which we permanently delete orders that have been flagged for removal. Previously we would wait for the cleanup job to handle this (once an hour), but that meant many removed orders would accumulate. We now prune them every 5 minutes. (#471)
- The
getStats
RPC endpoint now includes a new field which accounts for the number of orders that have been marked as "removed" but not yet permanently deleted (#461). - Improved historical order sharing using round-robin algorithm instead of random selection (#454). This will reduce the warm-up time for receiving existing orders when first joining the network.
- Added ERC1155 assetData support (#453). This includes order watching and order events for orders involving ERC1155 tokens.
- Added Ability to specify custom contract addresses via the
CUSTOM_ADDRESSES
environment variable or thecustomAddresses
field in the TypeScript bindings (#445).
- Temporarily disabled bandwidth-based peer banning (#448). A bug in libp2p was occasionally causing observed bandwidth usage to spike to unrealistic levels, which can result in peers being erroneously banned. We decided to temporarily stop banning peers while we're working with the libp2p team to resolve the issue.
- Updated DevUtils.sol contract address for the Kovan network to one including a bug fix for validating orders with nulled out
feeAssetData
fields. (#446]) - Added back Ropsten and Rinkeby support and fixed
exchange
address on Kovan (#451) - Segregate V3 Mesh network from V2 network (#455)
- Removes the
txHashes
key in theOrderEvent
s emitted from theorders
JSON-RPC subscription and replaced it withcontractEvents
, an array of decoded order-relevant contract events. Parsing these events allows callers to find every discrete order fill/cancel event. (#420) - Renames the
Kind
key inOrderEvent
toEndState
to better elucidate that it represents the aggregate change to the orders state since it was last re-validated. As an end state, it does not capture any possible intermediate states the order might have been in since the last re-validation. Intermediate states can be inferred from thecontractEvents
included (#420)
- Removed the max expiration limit for orders. The only remaining expiration constraint is that the unix timestamp does not overflow int64 (i.e., is not larger than 9223372036854775807). (#400)
- Fixed bug where we weren't updating an orders
fillableTakerAssetAmount
in the DB when orders were being partially filled or when their fillability increased due to a block re-org. (#439) - Made
verbosity
field optional in the TypeScriptConfig
type. (#410) - Fixed issue where we weren't re-validating orders potentially impacted by the balance increase of the recipient of an ERC20 or ERC721 transfer. (#416)
- Fixed a DB transaction deadlock accidentally introduced in the v4.0.0-beta release. (#403)
- Renamed the environment variable
P2P_LISTEN_PORT
toP2P_TCP_PORT
(#366). This makes it possible to configure Mesh to use both the TCP and Websocket transports by listening on different ports.
- Enabled WebSocket transport for bootstrap nodes and all other nodes (#361, #363, #366). By default the WebSocket transport listens on port
60559
but this can be changed via theP2P_WEBSOCKETS_PORT
environment variable. - Created TypeScript bindings and an NPM package for running Mesh directly in the browser (#369). Documentation for the NPM package and a guide for running Mesh in the browser can be found at https://0x-org.gitbook.io/mesh/.
- Added ability to use custom bootstrap list via the
BOOTSTRAP_LIST
environment variable (#374). Typically this should only be used for testing/debugging. - Added WebAssembly/Browser support to packages that previously did not support it (#358, #359, #366).
- Order hash calculations are now cached, which slightly improves performance (#365).
- Refactored
BlockWatch
so that it can be used without usingLevelDB
for Ethereum block storage. (#355)
- Fixed two related bugs: One where order expiration events would be emitted multiple times and another that meant subsequent fill/cancel events for orders deemed expired were not emitted. Fills/cancels for expired orders will continue to be emitted if they occur within ~4 mins (i.e. 20 blocks) of the expiration (#385).
- Fixed a data race-condition in OrderWatcher that could have caused order collection updates to be overwritten in the DB. (#386)
- Fixed a bug where
fillableTakerAssetAmount
andlastUpdated
were not always being properly updated in the DB. (#386) - Fixed some issues with key prefixes for certain types not being applied correctly to logs (#375).
- Fixed an issue where order hashes were not being correctly logged (#368).
- Mesh will now properly shut down if the database is unexpectedly closed (#370).
- Fixed bug where block number would sometimes be converted to hex with a leading zero, an invalid hex value per the Ethereum JSON-RPC specification. (#353)
- Fixed bug which resulted in orders that were close to expiring being re-added and removed multiple times, resulting in multiple ADDED and EXPIRED events for the same order (#352).
- Modified Mesh's validation logic to reject and consider invalid any partially fillable orders. While this is technically a breaking change, partially fillable orders are rare in the wild and we don't expect this will affect many users. (#333)
- Lowercased
GetStatsAsync
method togetStatsAsync
in TS client
- De-dup order submitted via the JSON-RPC method
mesh_addOrders
before performing validation (#331) - Added
"declaration": true,
to TS client'stsconfig.json
so that downstream projects can use it's TS typings. (#325)
- Modified how
mesh_addOrders
treats orders that are already stored on the Mesh node. Previously, they would be rejected with codeOrderAlreadyStored
. Now, if the order is stored and fillable, it will be accepted. If it is stored but unfillable, it will be rejected withOrderAlreadyStoredAndUnfillable
. We additionally added aisNew
property to the accepted orderInfos returned, so that callers can discern which orders Mesh already knew about. (#316)
- Added backup bootstrap nodes provided by the libp2p community
- Improved log formatting and reduced verbosity in a few cases (#314, #287)
- Reduced AdvertiseBootDelay for bootstrap nodes
- Implemented a check that will alerts you when switching to a different Ethereum network ID. (#301) -- special thanks to @hrharder!
- Made environment variable parsing more generous by automatically removing quotes if needed (#306)
- Improved tests by adding timeouts and closing resources where appropriate (#310, #309, #308)
- Increased robustness by removing panics and failing more gracefully (#312)
- RPC server is now started while block event backfilling is happening under the hood instead of waiting for it to complete (#318)
- Added a
mesh_getStats
endpoint which returns a host of useful information about the state of the Mesh node (e.g., number of fillable order stored, number of peers, peerID, etc...) (#322)
- Log messages are no longer incorrectly fired when receiving orders which have already been seen (#286)
- Fixed a bug where Mesh was still running after the database was closed (#300)
- Handled Parity "unknown block" error gracefully like we do Geth's (#285)
This release fixes several bugs:
- Uninitialized TxHashes map & accidental inclusion of null address txHash in order events (#280)
- Concurrent read/write issue in OrderWatcher's EventDecoder (#278)
- Non-unique logging keys causing Elastic Search indexing issues (#275)
It also includes a reduction in the delay before which bootstrap nodes advertise themselves as relays from 15mins to 30sec.
This version introduces a temporary hack in order to help some Mesh nodes find their public IP address under certain circumstances.
This release fixes a bug in how AutoNAT and AutoRelay services were discovered.
This release fixes some networking related issues with our bootstrap nodes.
- Bootstrap nodes now advertise the correct public IP address.
- Bootstrap nodes now also operate as a relay.
These fixes will help smooth out any issues with peer discovery.
This release fixes a few minor bugs and includes additional documentation.
- Set a custom protocol ID for our DHT in order to separate it from the default IPFS DHT.
- Fixed a bug in the getOrders JSON-RPC endpoint where fillableTakerAssetAmount was sometimes being encoded as numbers instead of strings.
- Converted addresses to all lowercase (non-checksummed) in the JSON-RPC API.
- Improved logging.
- Added a guide for running a Mesh node with telemetry-enabled.
This release adds AutoNAT support for our bootstrap nodes. This enables peers who are behind NATs to find each and connect to each other (in most cases).
In addition, this release includes some changes to default network timeouts and documentation improvements.
This is the initial beta release for 0x Mesh!
This release supports the following features:
- Automatic peer discovery via a DHT.
- A JSON-RPC endpoint for interacting with your Mesh node. It includes support for adding new orders and subscribing to order updates.
- Efficient order validation and order watching under the hood. You will get notified quickly when orders are expired, canceled, or filled.
- Basic limitations on the size and types of messages sent by peers. Peers that send malformed messages or messages that are too big will be dropped.
In addition to improving stability and scalability we plan to release many more important features in the near future. Check out https://github.com/0xProject/0x-mesh/issues for more information about what we are working on.