Skip to content

wallet: fix rescanning inconsistency

Placeholder Martin Zumsande requested to merge github/fork/mzumsande/202501_rescan_bestblock into master

If the chain advances during a rescan, ScanForWalletTransactions would previously process the new blocks without adjusting m_last_processed_block, which would leave the wallet in an inconsistent state temporarily, and could lead to crashes in the GUI reported in #31474. Fix this by not rescanning blocks beyond m_last_processed_block - for all blocks beyond that height, there will be pending BlockConnected notifications that will process them after the rescan is finished.

This means that if rescanning was triggered with cs_wallet permanently held (AttachChain), additional blocks that were connected during the rescan will only be processed with the pending blockConnected notifications after the lock is released. If rescanning without a permanent cs_wallet lock (RescanFromTime), additional blocks that were connected during the rescan can be re-processed here because m_last_processed_block was already updated by blockConnected.

Fixes #31474

Merge request reports