okhttp3 文件描述符问题太多

okhttp3 too many file descriptors issue

我有一张图片 "manager" 可以下载图片。 以前我使用毕加索库如下

class DownloadImage implements Runnable {

    String url;
    Context context;

    public DownloadImage(String url, Context context) {
        this.url = url;
        this.context = context;
    }

    @Override
    public void run() {
        try {
            String hash = Utilities.getSha1Hex(url);
            FileOutputStream fos = context.openFileOutput(hash, Context.MODE_PRIVATE);
            Bitmap bitmap = Picasso.with(context)
                    .load(url)
                    .resize(1024, 0) // Height 0 to ensure the image is scaled with respect to width - 
                    .onlyScaleDown()
                    .memoryPolicy(MemoryPolicy.NO_CACHE)
                    .get();

            // Writing the bitmap to the output stream
            bitmap.compress(Bitmap.CompressFormat.JPEG, 80, fos);
            fos.close();
            bitmap.recycle();
        } catch (IOException e) {
            Timber.e(e, "For url %s", url);
        } catch (OutOfMemoryError e) {
            Timber.e(e, "out of memory for url %s", url);
        }
    }
}

但这会创建一个 Bitmap 对象,它不仅会消耗大量内存,而且速度也相当慢且没有必要。

我已修改此 Runnable 以使用 okhttp3 代替:

class DownloadImage implements Runnable {

    String url;
    Context context;

    public DownloadImage(String url, Context context) {
        this.url = url;
        this.context = context;
    }

    @Override
    public void run() {
        try {
            String hash = Utilities.getSha1Hex(url);
            final FileOutputStream fos = context.openFileOutput(hash, Context.MODE_PRIVATE);
            Request request = new Request.Builder().url(url).build();
            okHttpClient.newCall(request).enqueue(new Callback() {
                @Override
                public void onFailure(Call call, IOException e) {
                    try {
                        fos.close();
                    } catch (IOException e1) {
                        e1.printStackTrace();
                    }
                }

                @Override
                public void onResponse(Call call, Response response) throws IOException {
                    Sink sink = null;
                    BufferedSource source = null;
                    try {
                        source = response.body().source();
                        sink = Okio.sink(fos);
                        source.readAll(sink);
                    } catch (Exception e) {
                        Timber.e(e, "Downloading an image went wrong");
                    } finally {
                        if (source != null) {
                            source.close();
                        }
                        if (sink != null) {
                            sink.close();
                        }
                        fos.close();
                        okHttpClient.connectionPool().evictAll(); // For testing
                    }
                }
            });
        } catch (IOException e) {
            Timber.e(e, "For url %s", url);
        }
    }
}

虽然这种方法比以前的方法快很多,但对于大量图像,我得到 A/libc: FORTIFY_SOURCE: FD_SET: file descriptor >= FD_SETSIZE. Calling abort(). 后跟一个微转储,这意味着我打开了太多文件描述符。

为了测试,我添加了 okHttpClient.connectionPool().evictAll(); // For testing 行,但是没有用。 我还尝试在构建 okHttpClient 时设置 builder.connectionPool(new ConnectionPool(4, 500, TimeUnit.MILLISECONDS));,但这也没有任何作用。 我也知道 https://github.com/square/okhttp/issues/2636

我好像都关闭了streams/sinks/sources,请问这是怎么回事?

使用 execute 函数将可运行对象添加到 ThreadPoolExecutor,创建方法如下:

// Sets the amount of time an idle thread waits before terminating
private static final int KEEP_ALIVE_TIME = 500;
// Sets the Time Unit to milliseconds
private static final TimeUnit KEEP_ALIVE_TIME_UNIT = TimeUnit.MILLISECONDS;
private static int NUMBER_OF_CORES = Runtime.getRuntime().availableProcessors();

// A queue of Runnables
private final BlockingQueue<Runnable> mDecodeWorkQueue;
private OkHttpClient okHttpClient;

ThreadPoolExecutor mDecodeThreadPool;

public ImageManager() {
    // Instantiates the queue of Runnables as a LinkedBlockingQueue
    mDecodeWorkQueue = new LinkedBlockingQueue<Runnable>();

    // Creates a thread pool manager
    mDecodeThreadPool = new ThreadPoolExecutor(
            NUMBER_OF_CORES,       // Initial pool size
            NUMBER_OF_CORES,       // Max pool size
            KEEP_ALIVE_TIME,
            KEEP_ALIVE_TIME_UNIT,
            mDecodeWorkQueue);

}

通过在 OnResponse 正文中创建和使用 FileOutputStream 解决了这个问题,因此在完成请求时它不会打开。