Releases: maticnetwork/bor
Releases · maticnetwork/bor
v0.3.4-beta3
Release notes (post v0.3.4-beta2):
Improvements and feature addition:
- Added a flag
rpc.batchlimit
to limit the maximum number of RPC requests in a batch. By default, 100 requests are allowed from this release otherwise the requests will be rejected (#657) - Added a flag
rpc.returndatalimit
to limit the maximum size of the result of an RPC request. By default, a maximum of 100000 bytes of result is allowed from this release otherwise the requests will be rejected (#657) - Fixed the misconfiguration of the
triesInMemory
flag. After this fix, a user can modify the number of block states (tries) stored in the memory (#691) - Added a flag
http.ep-size
andws.ep-size
to limit the execution pool size for HTTP and WS requests. Default value is 40 (#719) - Optimize
GetBorReceipt
function (#701) - Discard transactions which has size more than max code init size (#739)
Bug fixes:
v0.3.5-pprof
Event based pprof (#732) * feature * Save pprof to /tmp --------- Co-authored-by: Jerry <[email protected]>
v0.3.5-beta
Release Notes:
- Added a flag
rpc.batchlimit
to limit the maximum number of RPC requests in a batch. By default, 100 requests are allowed from this release otherwise the requests will be rejected. - Added a flag
rpc.returndatalimit
to limit the maximum size of the result of an RPC request. By default, a maximum of 100000 bytes of result is allowed from this release otherwise the requests will be rejected. - Fixed the misconfiguration of the triesInMemory flag. After this fix, a user can modify the number of block states ( tries ) stored in the memory.
- Added a flag
ws.ep-size
to limit the execution pool size for WS requests. Default value is 40 - Added a flag
http.ep-size
to limit the execution pool size for HTTP requests. Default value is 40 - Bug fixes for state sync
NewDeposits
- Optimize
GetBorReceipt
v0.3.3-beta3
Merge pull request #702 from maticnetwork/jerryc/temp-bor-sync Jerryc/temp bor sync
v0.3.4-beta2
Increase grpc message size limit in pprof
v0.3.3-beta2
Increase grpc message size limit in pprof
v0.3.4-beta
Release Notes:
- Added a flag
rpc.batchlimit
to limit the maximum number of RPC requests in a batch. By default, 100 requests are allowed from this release otherwise the requests will be rejected. - Added a flag
rpc.returndatalimit
to limit the maximum size of the result of an RPC request. By default, a maximum of 100000 bytes of result is allowed from this release otherwise the requests will be rejected. - Fixed the misconfiguration of the
triesInMemory
flag. After this fix, a user can modify the number of block states ( tries ) stored in the memory. - Fixed issue of huge rewinds.
trieTimeout
/trieTimeLimit
is responsible for the frequency of storing block state to the persistent db for a full node. It was set to 60 minutes which resulted in a huge gap between states of blocks in the persistent memory. We have reduced this gap to 10 minutes which will increase the frequency of storing the block states in the persistent database.
v0.3.3
Change log:
Hard fork specific changes:
- Decrease sprint length from 64 to 16
- Increase
baseFeeChangeDenominator
from 8 to 16
New-cli:
- Pruning has been introduced in the new-cli (using the command
bor snapshot prune-state
). The internal mechanism of pruning will work in the same way as it used to with the previous versions. The command snapshot prune-state will be available as a subcommand of the bor binary and will accept relevant flags (e.g. datadir). More details can be found here. - Adds support for a new flag for tweaking the number of tries in memory (
—cache.triesinmemory
, default: 128) (#552). - Updates and improvements in the description and documentation of the flags in new-cli along with their default values in the help command.
Improvements and feature additions:
- GRPC integration for bor and heimdall (#468).
- Improve bor’s interaction with heimdall by using context based cancellation methods in heimdall client and bor consensus (#445).
- More improvements on checkpoint based validations while importing new blocks to control reorgs (#425).
- Improvement and fixes in
eth_getTransactionReceiptsByBlock
rpc call which used to return wrong results when state-sync transactions are present and unit tests for the same (#477). - Replace mutex wherever possible with atomic pointers (introduced in go 1.19) (#446).
- Add new metrics in heimdall client to measure requests (like valid/invalid requests, duration of each request, etc) (#459).
- Add traces (available through open-telemetry tracing) in whole mining process to measure time taken by each component in building a new block to make improvements in future (#429).
- Enable tracing support for state-sync transactions in
debug
requests (#547). - Minor bug fixes and improvements across other modules.
Testing and benchmarking
- Restore geth tests (#436).
- Benchmark mining operations to make it more optimal to use (#439).
- More unit tests for testing the consistency of validator set operations (e.g. handling priority and power of validators in set) in bor consensus (#452).
- e2e tests for bor receipt calculations (#431).
- Unit tests for bor specific filters (#456).
- Unit tests for scenarios when validator comes back online after a crash and tries to reorg (#507).
- Property based tests on transaction pool have been added to make improvements in future (#466).
v0.3.2
Release notes:
- Decrease sprint length from 64 to 16
- Increase baseFeeChangeDenominator from 8 to 16
- GRPC integration for bor and heimdall
- Updates and improvements in the description and documentation of the flags in new-cli along with their default values in the help command.
- Pruning has been introduced in the new-cli. The internal mechanism of pruning will work in the same way as it used to with the previous versions. The command snapshot prune-state will be available as a subcommand of the bor binary and will accept relevant flags (e.g. --datadir). More details can be found here.
- Bug fixes and improvements
v0.3.2-beta
Bor v0.3.2-beta is a small patch release for the nodes running v0.3.1-mumbai on the Mumbai Testnet.
Change log
- Revert addition of actual
miner/author
address at the RPC layer foreth_getBlockByNumber
calls. It will now return “0x00” as the miner of the block as it used to do earlier. You can get the miner address of a block usingbor_getAuthor
call. This revert was done because started breaking downstream tooling/applications which rely on the equivalence ofCOINBASE
opcode and the miner field in response to the RPC request. The underlying miner value of the block (COINBASE) was still “0x00” and only the RPC layer had been changed earlier. - Updates and improvement in the description and documentation of the flags in new-cli along with their default values in the help command.
- Pruning has been introduced in the new-cli. The internal mechanism of pruning will work in the same way as it used to with the previous versions. The command
snapshot prune-state
will be available as a subcommand of the bor binary and will accept relevant flags (e.g.--datadir
). More details can be found here.