If you are using `response_model` with some type that doesn't include `None` but the function is returning `None`, it will now raise an internal server error, because you are returning invalid data that violates the contract in `response_model`. Before this release it would allow breaking that contract returning `None`.

For example, if you have an app like this:

```Python
from fastapi import FastAPI

  public static CharMatcher invisible() {
    return Invisible.INSTANCE;
  }

  /**
   * Determines whether a character is single-width (not double-width). When in doubt, this matcher
   * errs on the side of returning {@code false} (that is, it tends to assume a character is
   * double-width).
   *
   * <p><b>Note:</b> as the reference file evolves, we will modify this matcher to keep it up to
   * date.
   *

    TF_Operation* op;
    UseHelper(inputs, &op);
    return op;
  }

  // All the *Helper methods are used as a workaround for the restrictions that
  // one cannot call ASSERT_* methods in non-void-returning functions (when
  // exceptions are disabled during compilation)
  void UseHelper(const std::vector<TF_Output>& inputs, TF_Operation** op) {
    TF_OperationDescription* desc =

    inappropriate.

 *  New: `HttpUrl.get(String)` is an alternative to `HttpUrl.parse(String)` that throws an exception
    when the URL is malformed instead of returning null. Use this to avoid checking for null in
    situations where the input is known to be well-formed. We've also added `MediaType.get(String)`
    which is an exception-throwing alternative to `MediaType.parse(String)`.

	return err
}

// getClusterMetaInfo gets information of the current cluster and
// returns it.
// This is not a critical function, and it is allowed
// to fail with a ten seconds timeout, returning nil.
func getClusterMetaInfo(ctx context.Context) []byte {
	objectAPI := newObjectLayerFn()
	if objectAPI == nil {
		return nil
	}

	// Add a ten seconds timeout because getting profiling data

* Changed default scsi controller type in vSphere Cloud Provider ([#38426](https://github.com/kubernetes/kubernetes/pull/38426), [@abrarshivani](https://github.com/abrarshivani))
* Fixes API compatibility issue with empty lists incorrectly returning a null `items` field instead of an empty array. ([#39834](https://github.com/kubernetes/kubernetes/pull/39834), [@liggitt](https://github.com/liggitt))

found in the _cgo_export.h generated header, after any preambles
copied from the cgo input files. Functions with multiple
return values are mapped to functions returning a struct.

Not all Go types can be mapped to C types in a useful way.
Go struct types are not supported; use a C struct type.
Go array types are not supported; use a C pointer.

			}
		}
	}

	modTime := opts.MTime
	if opts.MTime.IsZero() {
		modTime = UTCNow()
	}
	fvID := mustGetUUID()

	defer func() {
		// attempt a heal before returning if there are offline disks
		// for both del marker and permanent delete situations.
		for _, disk := range storageDisks {
			if disk != nil && disk.IsOnline() {
				continue
			}

* Fix incorrectly propagated ResourceVersion in ListRequests returning 0 items. ([#64150](https://github.com/kubernetes/kubernetes/pull/64150), [@wojtek-t](https://github.com/wojtek-t))

protobuf you will be forced to restore from backup, and you will lose any changes
since you converted to protobuf.  After conversion to protobuf, you should
validate the correct operation of your cluster thoroughly before returning it
to normal operation.

Backups are always a good precaution, particularly around upgrades, but this
upgrade has multiple known issues where the only remedy is to restore from
backup.