java - Files.walkFileTree 使用自定义 FileVisitor 泄漏目录描述符

Question

我在我的应用程序中遇到了一个奇怪的错误。我已经通过解决方法解决了这个问题，但我仍然很好奇为什么会发生这个错误。

下面给出了一个自定义 FileVisitor 的示例，该示例正在删除它所遍历的空目录。如果目录不为空，并且它仍然遍历这些目录，则会泄漏目录描述符。如果我将 lsof 与应用程序的 PID 一起使用，它将显示一堆指向相同的几个目录的描述符，即它遍历的目录。

private String getOldestFile() {
    fileVisitor.clearOldestFile();

    try {
        // FIXME: this was throwing FileSystemException: Too many open files after some time running. Leaking file descriptors!!
        Files.walkFileTree(Paths.get(csvPath), fileVisitor);
    } catch (IOException e) {
        e.printStackTrace();
    }

    return fileVisitor.getOldestFile().toString();
}

class CustomFileVisitor extends SimpleFileVisitor<Path> {
    private Path oldestFile = null;

    Path getOldestFile() {
        return oldestFile;
    }

    void clearOldestFile() {
        oldestFile = null;
    }

    @Override
    public FileVisitResult visitFile(Path file, BasicFileAttributes attrs) throws IOException {
        if (attrs.isDirectory())
            return FileVisitResult.CONTINUE;

        if (oldestFile == null)
            oldestFile = file;

        if (oldestFile.compareTo(file) > 0)
            oldestFile = file;

        return FileVisitResult.CONTINUE;
    }

    @Override
    public FileVisitResult postVisitDirectory(Path dir, IOException exc) throws IOException {
        if (dir.equals(Paths.get(csvPath)))
            return FileVisitResult.CONTINUE;

        if (Files.list(dir).collect(Collectors.toList()).size() == 0)
            Files.delete(dir); // throws an exception if folder is not empty -> mustn't delete folder with files

        return FileVisitResult.CONTINUE;
    }
}

CustomFileVisitor 仅在外部类中创建一次，并且定期调用该函数，如 filename = getOldestFile();

编辑:发布lsof -p {PID}输出。一开始，我找到PID如this .

这就是 lsof -p {PID} 输出的样子，只是有数千行。 “/home/leon/Development/data/”是 Files.walkFileTree 的输入。

java    14965 leon  285r      DIR                8,2     4096  1970798 /home/leon/Development/data/2017
java    14965 leon  286r      DIR                8,2     4096  1970799 /home/leon/Development/data/2017/10
java    14965 leon  287r      DIR                8,2     4096  1970799 /home/leon/Development/data/2017/10
java    14965 leon  288r      DIR                8,2    36864  1970800 /home/leon/Development/data/2017/10/17
java    14965 leon  289r      DIR                8,2    36864  1970800 /home/leon/Development/data/2017/10/17
java    14965 leon  290r      DIR                8,2     4096  1970798 /home/leon/Development/data/2017
java    14965 leon  291r      DIR                8,2     4096  1970798 /home/leon/Development/data/2017
java    14965 leon  292r      DIR                8,2     4096  1970799 /home/leon/Development/data/2017/10
java    14965 leon  293r      DIR                8,2     4096  1970799 /home/leon/Development/data/2017/10
java    14965 leon  294r      DIR                8,2    36864  1970800 /home/leon/Development/data/2017/10/17
java    14965 leon  295r      DIR                8,2    36864  1970800 /home/leon/Development/data/2017/10/17

编辑2:我已设法将问题隔离到这一行:Files.list(dir).collect(Collectors.toList()).size() == 0。这不应该被垃圾收集吗？

Answer 1

来自Files#list() documentation :

The returned stream encapsulates a DirectoryStream. If timely disposal of file system resources is required, the try-with-resources construct should be used to ensure that the stream's close method is invoked after the stream operations are completed.

最终，该流将被垃圾收集，但不会立即进行。所以在这种情况下你必须自己管理。