* one of those methods.
   */
  open fun responseFailed(
    call: Call,
    ioe: IOException,
  ) {
  }

  /**
   * Invoked immediately after a call has completely ended.  This includes delayed consumption
   * of response body by the caller.
   *
   * This method is always invoked after [callStart].
   */
  open fun callEnd(call: Call) {
  }

  /**

				split[idx] = sh
				if idx >= k && len(sh) > 0 {
					hasParity++
				}
			}
			fmt.Printf("Have %d complete remapped data shards and %d complete parity shards. ", ok, hasParity)

			if err := rs.ReconstructData(split); err == nil {
				fmt.Println("Could reconstruct completely")
				for i, data := range split[:k] {
					start := i * len(data)
					copy(m.mapped[start:], data)
				}

The client will get a **500 Internal Server Error** for that request, but the application will continue working for the next requests instead of just crashing completely.

### Bigger Errors - Crashes { #bigger-errors-crashes }

Nevertheless, there might be cases where we write some code that **crashes the entire application** making Uvicorn and Python crash. 💥

      while (tableSize * DESIRED_LOAD_FACTOR < setSize) {
        tableSize <<= 1;
      }
      return tableSize;
    }

    // The table can't be completely full or we'll get infinite reprobes
    checkArgument(setSize < MAX_TABLE_SIZE, "collection too large");
    return MAX_TABLE_SIZE;
  }

  /**

					ModTime:   baseTime.Add(30 * time.Minute).UnixNano(),
					Signature: [4]byte{1, 1, 1, 1},
					Type:      ObjectType,
					Flags:     0,
				}},
			},
		},
		// Has a completely different version id
		3: {
			versions: []xlMetaV2ShallowVersion{
				{header: xlMetaV2VersionHeader{
					VersionID: [16]byte{3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1},

            socket = socket,
          ).buffer()
      requestBodySink.use {
        when {
          peek.doNotReadRequestBody -> {
            hasBody = false // Ignore the body completely.
          }

          contentLength != -1L -> {
            hasBody = contentLength > 0L || HttpMethod.permitsRequestBody(request.method)
            requestBodySink.write(socket.source, contentLength)

      while (tableSize * DESIRED_LOAD_FACTOR < setSize) {
        tableSize <<= 1;
      }
      return tableSize;
    }

    // The table can't be completely full or we'll get infinite reprobes
    checkArgument(setSize < MAX_TABLE_SIZE, "collection too large");
    return MAX_TABLE_SIZE;
  }

  /**

	// Make sure that the input format is GNU.
	// Unfortunately, the STAR format also has a sparse header format that uses
	// the same type flag but has a completely different layout.
	if blk.getFormat() != FormatGNU {
		return nil, ErrHeader
	}
	hdr.Format.mayOnlyBe(FormatGNU)

	var p parser
	hdr.Size = p.parseNumeric(blk.toGNU().realSize())
	if p.err != nil {

	if !ok {
		return "", fmt.Errorf("id_token not found!")
	}

	// fmt.Printf("TOKEN: %s\n", rawIDToken)
	return rawIDToken, nil
}

// unwrapAll will unwrap the returned error completely.
func unwrapAll(err error) error {
	for {
		werr := errors.Unwrap(err)
		if werr == nil {
			return err
		}
		err = werr
	}
}

# instead depend on the dummy plugins in //tensorflow/compiler/xla/tsl/platform/default/build_config
# and use header only targets.
# TODO(ddunleavy): This seems completely broken. :tensorflow_cc depends on
# cuda_platform from tf_additional_binary_deps and this doesn't break.
check_deps(
    name = "cuda_plugins_check_deps",
    disallowed_deps = if_static(
        [],