nf, err = tryReadFull(zeroReader{}, bf)
		}
		b = b[nf:]
		sr.pos += int64(nf)
		if sr.pos >= holeEnd && len(sr.sp) > 1 {
			sr.sp = sr.sp[1:] // Ensure last fragment always remains
		}
	}

	n = len(b0) - len(b)
	switch {
	case err == io.EOF:
		return n, errMissData // Less data in dense file than sparse file
	case err != nil:
		return n, err

	}
}

type panicReader struct{ panic bool }

func (r panicReader) Read(p []byte) (int, error) {
	if r.panic {
		panic("oops")
	}
	return 0, io.EOF
}

// Make sure that an empty Buffer remains empty when
// it is "grown" before a Read that panics
func TestReadFromPanicReader(t *testing.T) {

	// First verify non-panic behaviour
	var buf Buffer
	i, err := buf.ReadFrom(panicReader{})
	if err != nil {

			bf := b[:min(int64(len(b)), dataEnd-sw.pos)]
			nf, err = sw.fw.Write(bf)
		}
		b = b[nf:]
		sw.pos += int64(nf)
		if sw.pos >= dataEnd && len(sw.sp) > 1 {
			sw.sp = sw.sp[1:] // Ensure last fragment always remains
		}
	}

	n = len(b0) - len(b)
	switch {
	case err == ErrWriteTooLong:
		return n, errMissData // Not possible; implies bug in validation logic
	case err != nil:
		return n, err

   * <ul>
   *   <li>It is <i>consistent</i>: for any reference {@code x}, multiple invocations of {@code
   *       hash(x}} consistently return the same value provided {@code x} remains unchanged
   *       according to the definition of the equivalence. The hash need not remain consistent from
   *       one execution of an application to another execution of the same application.

 *
 * <p>The iteration ordering of the collections generated by {@code keySet}, {@code keys}, and
 * {@code asMap} has a few subtleties. As long as the set of keys remains unchanged, adding or
 * removing mappings does not affect the key iteration order. However, if you remove all values
 * associated with a key and then add the key back to the multimap, that key will come last in the

 * iteration ordering across rows and columns in the table's views. None of the view iterators
 * support {@link Iterator#remove}. If the table is modified after an iterator is created, the
 * iterator remains valid.
 *
 * <p>This class requires less memory than the {@link HashBasedTable} and {@link TreeBasedTable}
 * implementations, except when the table is sparse.
 *
 * <p>Null row keys or column keys are not permitted.
 *

- CAdvisor remains unchanged at v0.33.2. ([#76291](https://github.com/kubernetes/kubernetes/pull/76291))
- event-exporter remains unchanged at v0.2.5. ([#77815](https://github.com/kubernetes/kubernetes/pull/77815))
- ip-masq-agent remains unchanged at v2.4.1. ([#77844](https://github.com/kubernetes/kubernetes/pull/77844))

    is invoked, then you must make a good faith effort to ensure that,
    in the event an application does not supply such function or
    table, the facility still operates, and performs whatever part of
    its purpose remains meaningful.

    (For example, a function in a library to compute square roots has
    a purpose that is entirely well-defined independent of the
    application.  Therefore, Subsection 2d requires that any

            V entryValue = e.getValue();

            if (entryValue == null) {
              ++modCount;
              setValue(e, value);
              newCount = this.count; // count remains unchanged
              this.count = newCount; // write-volatile
              return null;
            } else if (onlyIfAbsent) {
              // Mimic
              // "if (!map.containsKey(key)) ...

- Es-image remains unchanged at Elasticsearch 6.3.2 since Kubernetes 1.12 ([#67484](https://github.com/kubernetes/kubernetes/pull/67484))
- metrics-server remains unchanged at v0.3.1 since Kubernetes 1.12 ([#68746](https://github.com/kubernetes/kubernetes/pull/68746))
- GLBC remains unchanged at v1.2.3 since Kubernetes 1.12 ([#66793](https://github.com/kubernetes/kubernetes/pull/66793))