internal/era: update block index format to be based on record offset (#28959)

As mentioned in #26621, the block index format for era1 is not in line with the regular era block index. This change modifies the index so all relative offsets are based against the beginning of the block index record.

cmd/utils/history_test.go

@@ -134,7 +134,7 @@ func TestHistoryImportAndExport(t *testing.T) {
 			for j := 0; it.Next(); j++ {
 				n := i*int(step) + j
 				if it.Error() != nil {
-					t.Fatalf("error reading block entry %d: %v", n, err)
+					t.Fatalf("error reading block entry %d: %v", n, it.Error())
 				}
 				block, receipts, err := it.BlockAndReceipts()
 				if err != nil {

internal/era/builder.go

@@ -49,7 +49,7 @@ import (
 //	CompressedBody     = { type: [0x04, 0x00], data: snappyFramed(rlp(body)) }
 //	CompressedReceipts = { type: [0x05, 0x00], data: snappyFramed(rlp(receipts)) }
 //	TotalDifficulty    = { type: [0x06, 0x00], data: uint256(header.total_difficulty) }
-//	Accumulator        = { type: [0x07, 0x00], data: accumulator-root }
+//	AccumulatorRoot    = { type: [0x07, 0x00], data: accumulator-root }
 //	BlockIndex         = { type: [0x32, 0x66], data: block-index }
 //
 // Accumulator is computed by constructing an SSZ list of header-records of length at most
@@ -64,8 +64,8 @@ import (
 //	block-index := starting-number | index | index | index ... | count
 //
 // starting-number is the first block number in the archive. Every index is a
-// defined relative to index's location in the file. The total number of block
+// defined relative to beginning of the record. The total number of block
-// entries in the file is recorded in count.
+// entries in the file is recorded with count.
 //
 // Due to the accumulator size limit of 8192, the maximum number of blocks in
 // an Era1 batch is also 8192.
@@ -115,12 +115,14 @@ func (b *Builder) Add(block *types.Block, receipts types.Receipts, td *big.Int)
 func (b *Builder) AddRLP(header, body, receipts []byte, number uint64, hash common.Hash, td, difficulty *big.Int) error {
 	// Write Era1 version entry before first block.
 	if b.startNum == nil {
-		if err := writeVersion(b.w); err != nil {
+		n, err := b.w.Write(TypeVersion, nil)
+		if err != nil {
 			return err
 		}
-		n := number
+		startNum := number
-		b.startNum = &n
+		b.startNum = &startNum
 		b.startTd = new(big.Int).Sub(td, difficulty)
+		b.written += n
 	}
 	if len(b.indexes) >= MaxEra1Size {
 		return fmt.Errorf("exceeds maximum batch size of %d", MaxEra1Size)
@@ -169,7 +171,7 @@ func (b *Builder) Finalize() (common.Hash, error) {
 		return common.Hash{}, fmt.Errorf("error writing accumulator: %w", err)
 	}
 	// Get beginning of index entry to calculate block relative offset.
-	base := int64(b.written + (3 * 8)) // skip e2store header (type, length) and start block
+	base := int64(b.written)
 
 	// Construct block index. Detailed format described in Builder
 	// documentation, but it is essentially encoded as:
@@ -186,7 +188,7 @@ func (b *Builder) Finalize() (common.Hash, error) {
 	// relative offset, the corresponding block can be quickly read by
 	// performing a seek relative to the current position.
 	for i, offset := range b.indexes {
-		relative := int64(offset) - (base + int64(i)*8)
+		relative := int64(offset) - base
 		binary.LittleEndian.PutUint64(index[8+i*8:], uint64(relative))
 	}
 	binary.LittleEndian.PutUint64(index[8+count*8:], uint64(count))
@@ -220,9 +222,3 @@ func (b *Builder) snappyWrite(typ uint16, in []byte) error {
 	}
 	return nil
 }
-
-// writeVersion writes a version entry to e2store.
-func writeVersion(w *e2store.Writer) error {
-	_, err := w.Write(TypeVersion, nil)
-	return err
-}

internal/era/era.go

@@ -221,9 +221,10 @@ func (e *Era) Count() uint64 {
 // is the absolute block number desired.
 func (e *Era) readOffset(n uint64) (int64, error) {
 	var (
-		firstIndex  = -8 - int64(e.m.count)*8               // size of count - index entries
+		blockIndexRecordOffset = e.m.length - 24 - int64(e.m.count)*8 // skips start, count, and header
-		indexOffset = int64(n-e.m.start) * 8                // desired index * size of indexes
+		firstIndex             = blockIndexRecordOffset + 16          // first index after header / start-num
-		offOffset   = e.m.length + firstIndex + indexOffset // offset of block offset
+		indexOffset            = int64(n-e.m.start) * 8               // desired index * size of indexes
+		offOffset              = firstIndex + indexOffset             // offset of block offset
 	)
 	e.mu.Lock()
 	defer e.mu.Unlock()
@@ -231,10 +232,10 @@ func (e *Era) readOffset(n uint64) (int64, error) {
 	if _, err := e.f.ReadAt(e.buf[:], offOffset); err != nil {
 		return 0, err
 	}
-	// Since the block offset is relative from its location + size of index
+	// Since the block offset is relative from the start of the block index record
-	// value (8), we need to add it to it's offset to get the block's
+	// we need to add the record offset to it's offset to get the block's absolute
-	// absolute offset.
+	// offset.
-	return offOffset + 8 + int64(binary.LittleEndian.Uint64(e.buf[:])), nil
+	return blockIndexRecordOffset + int64(binary.LittleEndian.Uint64(e.buf[:])), nil
 }
 
 // newReader returns a snappy.Reader for the e2store entry value at off.