Send trailing headers on a stream. Trailers are headers sent after the response body, commonly used for gRPC status codes, checksums,

Enable or disable accepting new connections.

Return overview of all processes (arbiter, web workers, dirty arbiter, dirty workers). Returns: Dictionary with complete

Return bound socket information. Returns: Dictionary with listeners list

Return list of HTTP workers. Returns: Dictionary with workers list containing: - pid: Worker process ID

Signal shutdown and wait for completion. This should be called during graceful shutdown.

Spawn a new dirty worker. Worker app assignment follows these priorities: 1. If there are pending respawns (from dead worker

\ Wait for signals with timeout. Returns a list of signals that were received.

Wake up the arbiter's main loop.

Increase worker count. Args: count: Number of workers to add (default 1) Returns: Dictionary with a

Gracefully terminate a specific worker. Args: pid: Worker process ID Returns: Dictionary with kille

Decrease worker count. Args: count: Number of workers to remove (default 1) Returns: Dictionary wit

Write a message to socket (sync). Args: sock: Socket to write to message: Message dict with 'type', 'id', an

Echo text messages.

Get size of buffered data.

Cancel the task if it's still pending after grace period.

Verify source IP is in the allowed list.

Check if peer address is in the trusted proxy list. Cached at connection start to avoid repeated IP parsing per request.

Close the connection.

Convert ASGI headers to HTTP/2 format (lowercase string names).

Create a LifespanManager with specific app.

Create a masked WebSocket frame. Args: fin: FIN bit (1 for final, 0 for continuation) opcode: Frame opcode

Create a mock HTTP request.

Create a mock HTTP request.

Create a mock HTTP request.

Create a mock HTTP request.

Create a mock HTTP request.

Create a WebSocketProtocol instance.

Create a WebSocketProtocol instance for testing.

Create a WebSocketProtocol instance.

Create an ASGIProtocol instance for testing.

Create an ASGIProtocol instance for testing.

Create an ASGIProtocol instance for testing.

Create an ASGIProtocol instance for testing.

Create an ASGIProtocol instance for testing.

Create an ASGIProtocol instance for testing.

Create an ASGIProtocol with worker state.

Create WebSocketProtocol with mock transport.

Extract HTTP request info from uWSGI vars. Header Mapping (CGI/WSGI to HTTP): The uWSGI protocol passes HTTP headers using CGI-style

Feed data to callback parser, handling parse errors. Returns True if parsing should continue, False if connection was closed.

Whether cleartext HTTP/2 could apply to this server at all.

Handle a single HTTP/2 request with streaming support. Streams both request and response body chunks immediately, avoiding buffering

Handle SIGCHLD - reap dead workers.

Handle WebSocket upgrade request.

Check if data contains only RFC 9110 token characters.

Simulate loading a model (expensive operation).

Callback: received body data chunk.

Parse uWSGI vars block into key-value pairs. Format: key_size (2 bytes LE) + key + val_size (2 bytes LE) + value

Whether the peer may set forwarding and scheme headers.

Pop a chunk and return the appropriate message.

Per RFC 9110 §6.4.2 a server MUST NOT send Content-Length on 1xx or 204 responses. HEAD and 304 are NOT covered: HEAD MAY include the

Restart server in parent after fork (called by fork handler).

Send a worker management message to the dirty arbiter. Args: operation: "add" or "remove" count: Number of w

Send HTTP response status and headers. Uses cached status lines and headers for common cases to avoid repeated string formatting and

Initiate async shutdown.

Start a consumer task for a worker's request queue.

Remove framing headers that must not appear on a no-body response. Transfer-Encoding is always stripped (chunked framing implies a body

Unregister a worker's apps when it exits. Removes the worker from all tracking maps. Args: worker_pid: The PID

Wait for gunicorn to be ready.

Benchmark arbiter message forwarding throughput.

Benchmark DirtyAsyncStreamIterator directly. Measures async iterator overhead vs raw message reading.

Compare stream() vs stream_async() performance with the same workload.

Benchmark time-to-first-chunk and time-to-last-chunk.

Benchmark worker streaming throughput with various chunk sizes.

A context manager that asserts post-conditions of ``listen_fds`` at exit. This helper is used to ease checking of the test post-conditions f

Close the client connection.

Build the (response, environ) pair for a request. ``response_class`` and ``response_args`` let a protocol supply its own writer: HTTP/2 passe

Create a test arbiter for benchmarks.

Create a worker instance for testing.

Create a worker instance for testing.

Create a worker for testing.

Create a worker for testing.

Create a worker for testing.

Create a worker for testing.

Create a worker for testing.

Create a worker for testing.

Create a worker for testing.

Create a worker for testing.