Client messages, which are always requests, are dispatched to Actor
instances one at a time via Actor::handle_message. Each request must be
paired with exactly one reply from the same actor the request was sent
to, where a reply is a message with no type (if a message from the
server has a type, it’s a notification, not a reply).
Failing to reply to a request will almost always permanently break that
actor, because either the client gets stuck waiting for a reply, or the
client receives the reply for a subsequent request as if it was the
reply for the current request. If an actor fails to reply to a request,
we want the dispatcher (ActorRegistry::handle_message) to send an error
of type `unrecognizedPacketType`, to keep the conversation for that
actor in sync. Since replies come in all shapes and sizes, we want to
allow Actor types to send replies without having to return them to the
dispatcher.
This patch adds a wrapper type around a client stream that guarantees
request/reply invariants. It allows the dispatcher to check if a valid
reply was sent, and guarantees that if the actor tries to send a reply,
it’s actually a valid reply (see ClientRequest::is_valid_reply). It does
not currently guarantee anything about messages sent via the TcpStream
released via ClientRequest::try_clone_stream or the return value of
ClientRequest::reply. We also send `unrecognizedPacketType`,
`missingParameter`, `badParameterType`, and `noSuchActor` messages per
the
[protocol](https://firefox-source-docs.mozilla.org/devtools/backend/protocol.html#error-packets)
[docs](https://firefox-source-docs.mozilla.org/devtools/backend/protocol.html#packets).
Testing: automated tests all pass, and manual testing looks ok
Fixes: #37683 and at least six bugs, plus one with a different root
cause, plus three with zero impact
---------
Signed-off-by: atbrakhi <atbrakhi@igalia.com>
Signed-off-by: Delan Azabani <dazabani@igalia.com>
Co-authored-by: delan azabani <dazabani@igalia.com>
Co-authored-by: Simon Wülker <simon.wuelker@arcor.de>
Co-authored-by: the6p4c <me@doggirl.gay>
* Use 2024 style edition
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
* Reformat all code
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
---------
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
For a long time, `gfx_traits` has held a lot of things unrelated to graphics
and also unrelated to the `gfx` crate (which is mostly about fonts).
This is a cleanup which does a few things:
1. Move non `gfx` crate things out of `gfx_traits`. This is important in
order to prevent dependency cycles with a different integration between
layout, script, and fonts.
2. Rename the `msg` crate to `base`. It didn't really contain anything
to do with messages and instead mostly holds ids, which are used
across many different crates in Servo. This new crate will hold the
*rare* data types that are widely used.
Details:
- All BackgroundHangMonitor-related things from base to a new
`background_hang_monitor_api` crate.
- Moved `TraversalDirection` to `script_traits`
- Moved `Epoch`-related things from `gfx_traits` to `base`.
- Moved `PrintTree` to base. This should be widely useful in Servo.
- Moved `WebrenderApi` from `base` to `webrender_traits` and renamed it
to `WebRenderFontApi`.
This does not (yet) upgrade ./rust-toolchain
The warnings:
* dead_code "field is never read"
* redundant_semicolons "unnecessary trailing semicolon"
* non_fmt_panic "panic message is not a string literal, this is no longer accepted in Rust 2021"
* unstable_name_collisions "a method with this name may be added to the standard library in the future"
* legacy_derive_helpers "derive helper attribute is used before it is introduced" https://github.com/rust-lang/rust/issues/79202
