CHECK(tensorflow::protobuf::TextFormat::ParseFromString(
      VariableAddFunctionSignature(), &def));
  return def.SerializeAsString();
}

// A graph optimization pass that would fail when triggered for more than once.
class GraphErrorInjectionPass : public tensorflow::GraphOptimizationPass {
 public:
  static bool enabled_;
  GraphErrorInjectionPass() {}

  absl::Status Run(

  open fun callFailed(
    call: Call,
    ioe: IOException,
  ) {
  }

  /**
   * Invoked when a call is canceled.
   *
   * Like all methods in this interface, this is invoked on the thread that triggered the event. But
   * while other events occur sequentially; cancels may occur concurrently with other events. For
   * example, thread A may be executing [responseBodyStart] while thread B executes [canceled].

    //
    // Except for (c), every thread should occupy the monitor very briefly, and every thread leaves
    // the monitor with the guard satisfied. Therefore as soon as tearDownLatch is triggered, we
    // should be able to enter the monitor, and then we set the guard to satisfied for the benefit
    // of any remaining waiting threads.

    tearDownLatch.countDown();
    assertTrue(

    //
    // Except for (c), every thread should occupy the monitor very briefly, and every thread leaves
    // the monitor with the guard satisfied. Therefore as soon as tearDownLatch is triggered, we
    // should be able to enter the monitor, and then we set the guard to satisfied for the benefit
    // of any remaining waiting threads.

    tearDownLatch.countDown();
    assertTrue(

      // otherwise have another task queued on to it. Note the exception to this, cancellation, is
      // specially handled in execute() - execute() calls triggered by cancellation are no-ops, and
      // thus don't count.
      requireNonNull(sequencer).latestTaskQueue = executingTaskQueue;
      sequencer = null;
      try {
        // requireNonNull is safe, as discussed above.

```Dockerfile
# ⛔️ Don't do this
CMD fastapi run app/main.py --port 80
```

Make sure to always use the **exec** form to ensure that FastAPI can shutdown gracefully and [lifespan events](../advanced/events.md) are triggered.

You can read more about it in the [Docker docs for shell and exec form](https://docs.docker.com/reference/dockerfile/#shell-and-exec-form).

    permitted. For example, a listener may receive `responseHeadersStart()` followed by
    `requestBodyEnd()`, both on the same call. Such events may be triggered by different threads
    even for a single call.

    Interceptors that rewrite or replace the request body may now inadvertently interfere with

   *       listeners can harm responsiveness, slow the system as a whole, or worse. (See also the
   *       note about locking below.)
   *   <li>If many tasks will be triggered by the same event, one heavyweight task may delay other
   *       tasks -- even tasks that are not themselves {@code directExecutor} tasks.
   *   <li>If many such tasks are chained together (such as with {@code

    module in 2018. We are confident in its stable API and solid implementation.
 *  Fix: Work around a crash in Android 10 and 11 that may be triggered when two threads
    concurrently close an SSL socket. This would have appeared in crash logs as
    `NullPointerException: bio == null`.
 *  Fix: Use plus `+` instead of `%20` to encode space characters in `FormBody`. This was a

- Fixed a 1.35.0 regression by disabling the `SchedulerAsyncAPICalls` feature gate. The regression was in scheduler performance, triggered by API client throttling. ([#135904](https://github.com/kubernetes/kubernetes/pull/135904), [@macsko](https://github.com/macsko)) [SIG Scheduling]