remove 2 debug_asserts

[jesup]

Remove a debug assert that could cause debug builds to fail in a race condition, plus one that should never happen

[Copilot]

Pull request overview

This PR removes two debug_assert!s in the extended CONNECT (WebTransport / CONNECT-UDP) session code to avoid debug-build failures in situations where those asserts can be reached (race/state transitions) or where the path should be unreachable.

Changes:

• WebTransportSession::write_datagram_capsule() now returns Err(Error::Unavailable) instead of unconditionally Ok(()) with a failing debug_assert!.
• Session::send_datagram() no longer fires a debug_assert!(false) when called in a non-Active state (it already returns Err(Error::Unavailable)).

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
neqo-http3/src/features/extended_connect/webtransport_session.rs	Replace an unconditional failing debug assert in the DATAGRAM capsule path with a concrete error return.
neqo-http3/src/features/extended_connect/session.rs	Remove a debug assert in send_datagram() that could trip during non-Active state transitions while preserving the existing error behavior.

[mxinden]

Can you explain how you are hitting this debug_assert?

webtransport::Session::datagram_capsule_support always returns false.

neqo/neqo-http3/src/features/extended_connect/webtransport_session.rs

Lines 402 to 413 in 6978589

	fn datagram_capsule_support(&self) -> bool {
	// HTTP/3 WebTransport requires QUIC datagram support. In other words,
	// HTTP/3 WebTransport never falls back to HTTP datagram capsules.
	//
	// > WebTransport over HTTP/3 also requires support for QUIC datagrams.
	// > To indicate support, both the client and the server send a
	// > max_datagram_frame_size transport parameter with a value greater than
	// > 0 (see Section 3 of [QUIC-DATAGRAM]).
	//
	// <https://www.ietf.org/archive/id/draft-ietf-webtrans-http3-14.html#section-3.1>
	false
	}

Thus webtransport::Session::write_datagram_capsule should enver be called.

neqo/neqo-http3/src/features/extended_connect/session.rs

Lines 449 to 457 in 6978589

	if conn.remote_datagram_size() == 0 && self.protocol.datagram_capsule_support() {
	qtrace!("[{self}] remote_datagram_size is 0, trying HTTP DATAGRAM Capsule");
	self.protocol.write_datagram_capsule(
	&mut self.control_stream_send,
	conn,
	buf,
	now,
	)?;
	return Ok((true, None));

[jesup]

Not hitting it. This shouldn't be possible at the moment, but if some future change (accidentally) changed that, returning an error would be better I think

[mxinden]

but if some future change (accidentally) changed that

In that case it would panic in debug mode, thus alerting us of the bug, and return an error in release mode, thus not causing harm. In other words, please keep this as is.

[mxinden]

👍

[mxinden]

@jesup please see my response above. In addition, I believe you accidentally included unrelated changes in this pull request.