Golang scanner.HashFile函数代码示例

// handleFile queues the copies and pulls as necessary for a single new or
// changed file.
func (f *rwFolder) handleFile(file protocol.FileInfo, copyChan chan<- copyBlocksState, finisherChan chan<- *sharedPullerState) {
	curFile, hasCurFile := f.model.CurrentFolderFile(f.folderID, file.Name)

	if hasCurFile && len(curFile.Blocks) == len(file.Blocks) && scanner.BlocksEqual(curFile.Blocks, file.Blocks) {
		// We are supposed to copy the entire file, and then fetch nothing. We
		// are only updating metadata, so we don't actually *need* to make the
		// copy.
		l.Debugln(f, "taking shortcut on", file.Name)

		events.Default.Log(events.ItemStarted, map[string]string{
			"folder": f.folderID,
			"item":   file.Name,
			"type":   "file",
			"action": "metadata",
		})

		f.queue.Done(file.Name)

		err := f.shortcutFile(file)
		events.Default.Log(events.ItemFinished, map[string]interface{}{
			"folder": f.folderID,
			"item":   file.Name,
			"error":  events.Error(err),
			"type":   "file",
			"action": "metadata",
		})

		if err != nil {
			l.Infoln("Puller: shortcut:", err)
			f.newError(file.Name, err)
		} else {
			f.dbUpdates <- dbUpdateJob{file, dbUpdateShortcutFile}
		}

		return
	}

	// Figure out the absolute filenames we need once and for all
	tempName, err := rootedJoinedPath(f.dir, defTempNamer.TempName(file.Name))
	if err != nil {
		f.newError(file.Name, err)
		return
	}
	realName, err := rootedJoinedPath(f.dir, file.Name)
	if err != nil {
		f.newError(file.Name, err)
		return
	}

	if hasCurFile && !curFile.IsDirectory() && !curFile.IsSymlink() {
		// Check that the file on disk is what we expect it to be according to
		// the database. If there's a mismatch here, there might be local
		// changes that we don't know about yet and we should scan before
		// touching the file. If we can't stat the file we'll just pull it.
		if info, err := f.mtimeFS.Lstat(realName); err == nil {
			if !info.ModTime().Equal(curFile.ModTime()) || info.Size() != curFile.Size {
				l.Debugln("file modified but not rescanned; not pulling:", realName)
				// Scan() is synchronous (i.e. blocks until the scan is
				// completed and returns an error), but a scan can't happen
				// while we're in the puller routine. Request the scan in the
				// background and it'll be handled when the current pulling
				// sweep is complete. As we do retries, we'll queue the scan
				// for this file up to ten times, but the last nine of those
				// scans will be cheap...
				go f.scan.Scan([]string{file.Name})
				return
			}
		}
	}

	scanner.PopulateOffsets(file.Blocks)

	var blocks []protocol.BlockInfo
	var blocksSize int64
	var reused []int32

	// Check for an old temporary file which might have some blocks we could
	// reuse.
	tempBlocks, err := scanner.HashFile(tempName, protocol.BlockSize, nil)
	if err == nil {
		// Check for any reusable blocks in the temp file
		tempCopyBlocks, _ := scanner.BlockDiff(tempBlocks, file.Blocks)

		// block.String() returns a string unique to the block
		existingBlocks := make(map[string]struct{}, len(tempCopyBlocks))
		for _, block := range tempCopyBlocks {
			existingBlocks[block.String()] = struct{}{}
		}

		// Since the blocks are already there, we don't need to get them.
		for i, block := range file.Blocks {
			_, ok := existingBlocks[block.String()]
			if !ok {
				blocks = append(blocks, block)
				blocksSize += int64(block.Size)
			} else {
				reused = append(reused, int32(i))
			}
//.........这里部分代码省略.........

// handleFile queues the copies and pulls as necessary for a single new or
// changed file.
func (p *rwFolder) handleFile(file protocol.FileInfo, copyChan chan<- copyBlocksState, finisherChan chan<- *sharedPullerState) {
	curFile, ok := p.model.CurrentFolderFile(p.folder, file.Name)

	if ok && len(curFile.Blocks) == len(file.Blocks) && scanner.BlocksEqual(curFile.Blocks, file.Blocks) {
		// We are supposed to copy the entire file, and then fetch nothing. We
		// are only updating metadata, so we don't actually *need* to make the
		// copy.
		if debug {
			l.Debugln(p, "taking shortcut on", file.Name)
		}

		events.Default.Log(events.ItemStarted, map[string]string{
			"folder": p.folder,
			"item":   file.Name,
			"type":   "file",
			"action": "metadata",
		})

		p.queue.Done(file.Name)

		var err error
		if file.IsSymlink() {
			err = p.shortcutSymlink(file)
		} else {
			err = p.shortcutFile(file)
		}

		events.Default.Log(events.ItemFinished, map[string]interface{}{
			"folder": p.folder,
			"item":   file.Name,
			"error":  events.Error(err),
			"type":   "file",
			"action": "metadata",
		})

		if err != nil {
			l.Infoln("Puller: shortcut:", err)
			p.newError(file.Name, err)
		} else {
			p.dbUpdates <- dbUpdateJob{file, dbUpdateShortcutFile}
		}

		return
	}

	events.Default.Log(events.ItemStarted, map[string]string{
		"folder": p.folder,
		"item":   file.Name,
		"type":   "file",
		"action": "update",
	})

	scanner.PopulateOffsets(file.Blocks)

	// Figure out the absolute filenames we need once and for all
	tempName := filepath.Join(p.dir, defTempNamer.TempName(file.Name))
	realName := filepath.Join(p.dir, file.Name)

	reused := 0
	var blocks []protocol.BlockInfo

	// Check for an old temporary file which might have some blocks we could
	// reuse.
	tempBlocks, err := scanner.HashFile(tempName, protocol.BlockSize)
	if err == nil {
		// Check for any reusable blocks in the temp file
		tempCopyBlocks, _ := scanner.BlockDiff(tempBlocks, file.Blocks)

		// block.String() returns a string unique to the block
		existingBlocks := make(map[string]struct{}, len(tempCopyBlocks))
		for _, block := range tempCopyBlocks {
			existingBlocks[block.String()] = struct{}{}
		}

		// Since the blocks are already there, we don't need to get them.
		for _, block := range file.Blocks {
			_, ok := existingBlocks[block.String()]
			if !ok {
				blocks = append(blocks, block)
			}
		}

		// The sharedpullerstate will know which flags to use when opening the
		// temp file depending if we are reusing any blocks or not.
		reused = len(file.Blocks) - len(blocks)
		if reused == 0 {
			// Otherwise, discard the file ourselves in order for the
			// sharedpuller not to panic when it fails to exclusively create a
			// file which already exists
			os.Remove(tempName)
		}
	} else {
		blocks = file.Blocks
	}

	s := sharedPullerState{
		file:        file,
		folder:      p.folder,
//.........这里部分代码省略.........

func TestCopierFinder(t *testing.T) {
	// After diff between required and existing we should:
	// Copy: 1, 2, 3, 4, 6, 7, 8
	// Since there is no existing file, nor a temp file

	// After dropping out blocks found locally:
	// Pull: 1, 5, 6, 8

	tempFile := filepath.Join("testdata", defTempNamer.TempName("file2"))
	err := os.Remove(tempFile)
	if err != nil && !os.IsNotExist(err) {
		t.Error(err)
	}

	// Create existing file
	existingFile := protocol.FileInfo{
		Name:     defTempNamer.TempName("file"),
		Flags:    0,
		Modified: 0,
		Blocks: []protocol.BlockInfo{
			blocks[0], blocks[2], blocks[3], blocks[4],
			blocks[0], blocks[0], blocks[7], blocks[0],
		},
	}

	// Create target file
	requiredFile := existingFile
	requiredFile.Blocks = blocks[1:]
	requiredFile.Name = "file2"

	db, _ := leveldb.Open(storage.NewMemStorage(), nil)
	m := NewModel(defaultConfig, protocol.LocalDeviceID, "device", "syncthing", "dev", db)
	m.AddFolder(defaultFolderConfig)
	// Update index
	m.updateLocals("default", []protocol.FileInfo{existingFile})

	iterFn := func(folder, file string, index int32) bool {
		return true
	}

	// Verify that the blocks we say exist on file, really exist in the db.
	for _, idx := range []int{2, 3, 4, 7} {
		if m.finder.Iterate(folders, blocks[idx].Hash, iterFn) == false {
			t.Error("Didn't find block")
		}
	}

	p := rwFolder{
		folder:    "default",
		dir:       "testdata",
		model:     m,
		errors:    make(map[string]string),
		errorsMut: sync.NewMutex(),
	}

	copyChan := make(chan copyBlocksState)
	pullChan := make(chan pullBlockState, 4)
	finisherChan := make(chan *sharedPullerState, 1)

	// Run a single fetcher routine
	go p.copierRoutine(copyChan, pullChan, finisherChan)

	p.handleFile(requiredFile, copyChan, finisherChan)

	pulls := []pullBlockState{<-pullChan, <-pullChan, <-pullChan, <-pullChan}
	finish := <-finisherChan

	select {
	case <-pullChan:
		t.Fatal("Finisher channel has data to be read")
	case <-finisherChan:
		t.Fatal("Finisher channel has data to be read")
	default:
	}

	// Verify that the right blocks went into the pull list
	for i, eq := range []int{1, 5, 6, 8} {
		if string(pulls[i].block.Hash) != string(blocks[eq].Hash) {
			t.Errorf("Block %d mismatch: %s != %s", eq, pulls[i].block.String(), blocks[eq].String())
		}
		if string(finish.file.Blocks[eq-1].Hash) != string(blocks[eq].Hash) {
			t.Errorf("Block %d mismatch: %s != %s", eq, finish.file.Blocks[eq-1].String(), blocks[eq].String())
		}
	}

	// Verify that the fetched blocks have actually been written to the temp file
	blks, err := scanner.HashFile(tempFile, protocol.BlockSize, 0, nil)
	if err != nil {
		t.Log(err)
	}

	for _, eq := range []int{2, 3, 4, 7} {
		if string(blks[eq-1].Hash) != string(blocks[eq].Hash) {
			t.Errorf("Block %d mismatch: %s != %s", eq, blks[eq-1].String(), blocks[eq].String())
		}
	}
	finish.fd.Close()

	os.Remove(tempFile)
}