Implement ra_proc_macro client logic #3738
Conversation
crates/ra_proc_macro/src/msg.rs
Outdated
| return Ok(None); | ||
| } | ||
| if !buf.ends_with("\r\n") { | ||
| return Err(invalid_data!("malformed header: {:?}", buf)); |
There was a problem hiding this comment.
Should we consider using anyhow for error-handling?
There was a problem hiding this comment.
Same as below, I just copy and paste from https://github.com/rust-analyzer/lsp-server/blob/master/src/msg.rs , but I am okay to change it if you want.
edwin0cheng
force-pushed
the
add-proc-macro-client-logic
branch
from
617f2cd to
26f7270
Compare
In general, I don't think we can, since we have no information about what structure the proc macro expects :( For attribute proc macros in particular though, we should probably try to 'complete' erroneous code before passing it to the proc macro somehow. At least in easy cases. |
edwin0cheng
force-pushed
the
add-proc-macro-client-logic
branch
from
d8c859d to
799bcba
Compare
|
Conflicts fixed. |
edwin0cheng
force-pushed
the
add-proc-macro-client-logic
branch
from
799bcba to
0f48ab4
Compare
crates/ra_proc_macro/src/process.rs
Outdated
| code: ErrorCode::ServerErrorEnd, | ||
| message: "Server closed".into(), | ||
| }); | ||
| if result_tx.send(res.into()).is_err() { |
There was a problem hiding this comment.
Similarly, lets make this an unwrap, and make sure that the opposite side of the channel is not dropped to early
There was a problem hiding this comment.
Fixed.
The recv side of this channel would not be dropped as it was passed by the task channel:
|
lets land this, so that it doesn't conflict with other stuff... bors r+ |
3738: Implement ra_proc_macro client logic r=matklad a=edwin0cheng This PR add the actual client logic for `ra_proc_macro` crate: 1. Define all necessary rpc serialization data structure, which include `ra_tt` related data and some task messages. Although adding `Serialize` and `Deserialize` trait to ra_tt directly seem to be much easier, we deliberately duplicate the `ra_tt` struct with `#[serde(with = "XXDef")]` for separation of code responsibility. 2. Define a simplified version of lsp base protocol for rpc, which basically copy from lsp-server code base. 3. Implement the actual `IO` for the client side progress spawning and message passing. Co-authored-by: Edwin Cheng <edwin0cheng@gmail.com>
|
bors r- |
Canceled |
|
ouch, sorry, this might now actually conflict with #3790, which I intended to land after this PR, but messed up :( Anyway, r=me |
edwin0cheng
force-pushed
the
add-proc-macro-client-logic
branch
from
94a068b to
3bc1670
Compare
|
bors r=matklad |
3738: Implement ra_proc_macro client logic r=matklad a=edwin0cheng This PR add the actual client logic for `ra_proc_macro` crate: 1. Define all necessary rpc serialization data structure, which include `ra_tt` related data and some task messages. Although adding `Serialize` and `Deserialize` trait to ra_tt directly seem to be much easier, we deliberately duplicate the `ra_tt` struct with `#[serde(with = "XXDef")]` for separation of code responsibility. 2. Define a simplified version of lsp base protocol for rpc, which basically copy from lsp-server code base. 3. Implement the actual `IO` for the client side progress spawning and message passing. Co-authored-by: Edwin Cheng <edwin0cheng@gmail.com>
Build failed |
|
bors retry |
Build succeeded |
4 participants
This PR add the actual client logic for
ra_proc_macrocrate:ra_ttrelated data and some task messages. Although addingSerializeandDeserializetrait to ra_tt directly seem to be much easier, we deliberately duplicate thera_ttstruct with#[serde(with = "XXDef")]for separation of code responsibility.IOfor the client side progress spawning and message passing.