silkworm: recreate tx batch and disable read-ahead in block execution #9293
Conversation
eth/stagedsync/stage_execute.go
Outdated
| @@ -468,6 +468,16 @@ Loop: | |||
| _, isMemoryMutation := txc.Tx.(*membatchwithdb.MemoryMutation) | |||
| if cfg.silkworm != nil && !isMemoryMutation { | |||
| blockNum, err = silkworm.ExecuteBlocks(cfg.silkworm, txc.Tx, cfg.chainConfig.ChainID, blockNum, to, uint64(cfg.batchSize), writeChangeSets, writeReceipts, writeCallTraces) | |||
| // Recreate tx because Silkworm has just done commit or abort on passed one | |||
| tx, tx_err := cfg.db.BeginRw(context.Background()) | |||
There was a problem hiding this comment.
-
probably you need
=instead of:=to assign on existingtxvariable (you doing it withbatch). -
It's bad pattern: to open tx on one layer of abstraction and close/commit in another. In this case high-level code can't provide "committed all or nothing" guarantee.
Better: if tx was created on top-level - then it must be closed/commit only at top-level, if on low-level - then close/commit on low-level.
We already using both (seeuseExternalTxvariable) - but not mixing them.
There was a problem hiding this comment.
- OK, I'm going to change it in this PR
- OK, I'll do it in a separate PR
here you mean when |
No description provided.