Incorrect progress report usage in BlobPuller

[chanseokoh]

From #1512 (comment).

  public Void handleResponse(Response response) ... {
    // Notifies the progress dispatcher that the total bytes we will read is the "Content-Length" value.
    // If `Content-Encoding: gzip`, the value will be the size of the compressed content.
    blobSizeConsumer.accept(response.getContentLength());

    try (OutputStream outputStream =
        new NotifyingOutputStream(destinationOutputStream, writtenByteCountListener)) {
      BlobDescriptor receivedBlobDescriptor =
          Digests.computeDigest(response.getBody(), outputStream);
      ...
    }

Now, in the case of Content-Encoding: gzip (or something similar), Google HTTP Client's HttpResponse::getContent() creates and returns GZIPInputStream that wraps the raw InputStream, which means the library does the chore of streamed unzipping on behalf of us. Content-Length will of course be the size of the original (compressed) content stream.

Then NotifyingOutputStream::write() will periodically notify the progress dispatcher how many bytes were written (through writtenByteCountListener). The problem is that the progress dispatcher is initialized with the total allocation size (through blobSizeConsumer) to be the value reported in Content-Length. Because we will write uncompressed content to outputStream in this case, we will go over the progress limit.

---

Another case where the progress is under reported:

jib/jib-core/src/main/java/com/google/cloud/tools/jib/builder/steps/ExtractTarStep.java

Line 155 in 85bb1c5

.create("compressing " + file, Files.size(file));

The NotifyingOutputStream will report written byte count. The count will be smaller than the original file size, as the written content is compressed.