// It returns the index where bytes should be written.
// If the buffer can't grow it will panic with ErrTooLarge.
func (b *Buffer) grow(n int) int {
	m := b.Len()
	// If buffer is empty, reset to recover space.
	if m == 0 && b.off != 0 {
		b.Reset()
	}
	// Try to grow by means of a reslice.
	if i, ok := b.tryGrowByReslice(n); ok {
		return i
	}
	if b.buf == nil && n <= smallBufferSize {

    assertThrows(IllegalStateException.class, () -> peekingIterator.remove());

    assertEquals(
        "After remove() throws exception, peek should still be ok", "B", peekingIterator.peek());

    /* Should recover to be able to remove() after next(). */
    assertEquals("B", peekingIterator.next());
    peekingIterator.remove();
    assertEquals("Should have removed an element", 2, list.size());

    assertThrows(IllegalStateException.class, () -> peekingIterator.remove());

    assertEquals(
        "After remove() throws exception, peek should still be ok", "B", peekingIterator.peek());

    /* Should recover to be able to remove() after next(). */
    assertEquals("B", peekingIterator.next());
    peekingIterator.remove();
    assertEquals("Should have removed an element", 2, list.size());

   * {@link TypeVariableImpl#equals(Object)} is changed to no longer be equal to custom TypeVariable
   * implementations. As a result, we need to make sure our TypeVariable implementation respects
   * symmetry. Moreover, we don't want to reconstruct a native type variable {@code <A>} using our
   * implementation unless some of its bounds have changed in resolution. This avoids creating

```
eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzdWIiOiIxMjM0NTY3ODkwIiwibmFtZSI6IkpvaG4gRG9lIiwiaWF0IjoxNTE2MjM5MDIyfQ.SflKxwRJSMeKKF2QT4fwpMeJf36POk6yJV_adQssw5c
```

It is not encrypted, so, anyone could recover the information from the contents.

But it's signed. So, when you receive a token that you emitted, you can verify that you actually emitted it.

    per_device_output_tensors.push_back(device_thread->Join(status));
    // We will run every Join even if there are bad statuses in case the user
    // wants to recover and continue running ops on the parallel device (which
    // would otherwise deadlock).
    if (TF_GetCode(status) != TF_OK &&
        (first_bad_status == nullptr

  // and not for correctness. The only downside of keeping this 1 seems to be
  // that the gradient accumulation is unbounded and we will never
  // aggressively aggregate accumulated gradients to recover memory.
  // Revisit and fix.
  return 1;
}

// Consumes references to the tensors in the gradient_tensors list and returns
// a tensor with the result.
AbstractTensorHandle* TapeVSpace::AggregateGradients(

</pre>

<h3 id="Handling_panics">Handling panics</h3>

<p> Two built-in functions, <code>panic</code> and <code>recover</code>,
assist in reporting and handling <a href="#Run_time_panics">run-time panics</a>
and program-defined error conditions.
</p>

<pre class="grammar">
func panic(interface{})
func recover() interface{}
</pre>

<p>
While executing a function <code>F</code>,

	// should panic with a description pointing at the reader, not at itself.
	// (should NOT panic with slice index error, for example.)
	b := NewReader(new(negativeReader))
	defer func() {
		switch err := recover().(type) {
		case nil:
			t.Fatal("read did not panic")
		case error:
			if !strings.Contains(err.Error(), "reader returned negative count from Read") {
				t.Fatalf("wrong panic: %v", err)
			}

				} else {
					err = errFileNotFound
				}
			}
		}
		// when we have insufficient read quorum and inconsistent metadata return
		// file not found, since we can't possibly have a way to recover this object
		// anyway.
		if v, ok := err.(InsufficientReadQuorum); ok && v.Type == RQInconsistentMeta {
			if opts.VersionID != "" {
				err = errFileVersionNotFound
			} else {
				err = errFileNotFound
			}