Commits on Aug 25, 2021

1. core/rawdb: batched freezer writes …

   core/rawdb: improve append benchmark
   
   core/rawdb: add freezer batch
   
   core/rawdb: minor benchmark nitpicks
   
   core/rawdb: add freezer batch
   
   core/rawdb: method docs
   
   core/rawdb: resolve freezer batch filenum at write-time
   
   core/rawdb: use bytes.Buffer in freezer batch
   
   core/rawdb: work in progress -- rlp encoding inside freezer batch
   
   core/rawdb: implement write batchs in freezer
   
   core/rawdb: improve memory allocations for snappy compression
   core/rawdb: polish
   core/rawdb: simplify + fix copy/paste error
   core/rawdb: error handling for freezertable batches
   core/rawdb: lint nitpicks
   core/rawdb: more tests for freezer batching
   
   core/rawdb: freezer batch interface
   
   core/rawdb: WIP freezer batch interface
   core/rawdb: check insert count of table batches
   core/rawdb: use ancient batch in background freeze
   core/rawdb: fix some issues in background freezer
   ethdb: document AncientBatch
   core/rawdb: remove single-item append
   core/rawdb: use dumpIndexString in batch tests
   core/rawdb: fix batch test
   core/rawdb: implement commit during append
   core/rawdb: add some missing close calls in freezer tests
   core/rawdb: simplify snappy buffer
   core/rawdb: change ancient writer interface
   core/rawdb: fix offset test
   core/rawdb: track headBytes only in freezerTable
   core/rawdb: add writeLock
   core/rawdb: reuse write batch
   core/rawdb: re-add metrics reporting
   core/rawdb: recreate the concurrent truncate test of freezer
   
   This changes the test slightly to use the 'freezer' object instead of
   'freezerTable'. This is necessary because the concurrency handling has
   moved to the freezer and concurrent append and truncate is no longer
   allowed on the table object.
   
   core/rawdb: fix item count after ModifyAncients and improve the test
   core/rawdb: add concurrency test for retrieve
   core/rawdb: allow overriding max table size in newFreezer
   
   This makes the concurrency tests fail with -race.
   
   core/rawdb: fix race in advanceFile
   core/rawdb: remove atomic access on freezer.headId
   
   It doesn't need to be atomic, all accesses are protected by the lock.
   
   core/rawdb: use int64 for writeSize
   
   core/rawdb: implement and test all-or-nothing behavior of ModifyAncients
   core/rawdb: move buffer code to batch file
   core/rawdb: reimplement batch test on freezer
   core/rawdb: fix race in AncientSize
   core, core/rawdb: WIP batch ancient write in sync
   core/rawdb: delete old batch tests
   core: fix error handling when inserting side chain receipts
   
   This fixes (well, rewrites) the test for the error case of
   InsertReceiptChain where side chain data is imported and setting the
   fast block fails because it doesn't match the header chain.
   
   This previously relied on the terminateInsert hook. With the new ancient
   write interface, there is no good place for this hook anymore, so the
   test now injects an actual error instead of simulating one.
   
   core: improve comments
   
   core/blockchain, core/rawdb: less iterator usage when deleting leveldb-data
   
   core/rawdb: return write error from WriteAncientBlocks
   
   It's easier to test this function when it doesn't just exit
   the process on failure.
   
   core/rawdb: add benchmark for WriteAncientBlocks
   
   core/rawdb: use os.RemoveAll in test
   
   core/rawdb: avoid allocating in benchmark loop
   
   core/rawdb: avoid copying difficulty twice
   
   The difficulty is already copied by block.Header(), avoid
   copying it again.
   
   core/rawdb: improve WriteAncientBlocks benchmark
   
   core: handle block->header validation

   

   holiman committed Aug 25, 2021
   Copy the full SHA

   1c473ed View commit details

   Browse the repository at this point in the history

2. core/rawdb: fix tests

   

   holiman committed Aug 25, 2021
   Copy the full SHA

   4d4cdf3 View commit details

   Browse the repository at this point in the history