Rollup merge of rust-lang#108846 - celinval:smir-poc, r=oli-obk

matthiaskrgr · web-flow · commit 693d05bf0659 · 2023-03-08T07:14:46.000+01:00
StableMIR: Proof-of-concept implementation + test This PR is part of the [project Stable MIR](https://github.com/rust-lang/project-stable-mir). The PR deletes old re-exports from rustc_smir and introduces a proof-of-concept implementation for APIs to retrieve crate information. The implementation follows the [design described here](https://hackmd.io/XhnYHKKuR6-LChhobvlT-g?view), but instead of using separate crates for the implementation, it uses separate modules inside `rustc_smir`. The API introduced at this point should be seen just as an example on how we are planning to structure the communication between tools and the compiler. I have not explored yet what should be the right granularity, the best starting point for users, neither the best way to implement it. r? ``@oli-obk``
diff --git a/Cargo.lock b/Cargo.lock
@@ -4669,15 +4669,9 @@ dependencies = [
 name = "rustc_smir"
 version = "0.0.0"
 dependencies = [
- "rustc_borrowck",
- "rustc_driver",
- "rustc_hir",
- "rustc_interface",
  "rustc_middle",
- "rustc_mir_dataflow",
- "rustc_mir_transform",
- "rustc_serialize",
- "rustc_trait_selection",
+ "rustc_span",
+ "tracing",
 ]
 
 [[package]]
diff --git a/compiler/rustc_smir/Cargo.toml b/compiler/rustc_smir/Cargo.toml
@@ -4,25 +4,12 @@ version = "0.0.0"
 edition = "2021"
 
 [dependencies]
-rustc_borrowck = { path = "../rustc_borrowck", optional = true }
-rustc_driver = { path = "../rustc_driver", optional = true }
-rustc_hir = { path = "../rustc_hir", optional = true }
-rustc_interface = { path = "../rustc_interface", optional = true }
 rustc_middle = { path = "../rustc_middle", optional = true }
-rustc_mir_dataflow = { path = "../rustc_mir_dataflow", optional = true }
-rustc_mir_transform = { path = "../rustc_mir_transform", optional = true }
-rustc_serialize = { path = "../rustc_serialize", optional = true }
-rustc_trait_selection = { path = "../rustc_trait_selection", optional = true }
+rustc_span = { path = "../rustc_span", optional = true }
+tracing = "0.1"
 
 [features]
 default = [
-    "rustc_borrowck",
-    "rustc_driver",
-    "rustc_hir",
-    "rustc_interface",
     "rustc_middle",
-    "rustc_mir_dataflow",
-    "rustc_mir_transform",
-    "rustc_serialize",
-    "rustc_trait_selection",
+    "rustc_span",
 ]
diff --git a/compiler/rustc_smir/README.md b/compiler/rustc_smir/README.md
@@ -73,3 +73,40 @@ git subtree pull --prefix=compiler/rustc_smir https://github.com/rust-lang/proje
 Note: only ever sync to rustc from the project-stable-mir's `smir` branch. Do not sync with your own forks.
 
 Then open a PR against rustc just like a regular PR.
+
+## Stable MIR Design
+
+The stable-mir will follow a similar approach to proc-macro2. It’s
+implementation will eventually be broken down into two main crates:
+
+- `stable_mir`: Public crate, to be published on crates.io, which will contain
+the stable data structure as well as proxy APIs to make calls to the
+compiler.
+- `rustc_smir`: The compiler crate that will translate from internal MIR to
+SMIR. This crate will also implement APIs that will be invoked by
+stable-mir to query the compiler for more information.
+
+This will help tools to communicate with the rust compiler via stable APIs. Tools will depend on
+`stable_mir` crate, which will invoke the compiler using APIs defined in `rustc_smir`. I.e.:
+
+```
+    ┌──────────────────────────────────┐           ┌──────────────────────────────────┐
+    │   External Tool     ┌──────────┐ │           │ ┌──────────┐   Rust Compiler     │
+    │                     │          │ │           │ │          │                     │
+    │                     │stable_mir| │           │ │rustc_smir│                     │
+    │                     │          │ ├──────────►| │          │                     │
+    │                     │          │ │◄──────────┤ │          │                     │
+    │                     │          │ │           │ │          │                     │
+    │                     │          │ │           │ │          │                     │
+    │                     └──────────┘ │           │ └──────────┘                     │
+    └──────────────────────────────────┘           └──────────────────────────────────┘
+```
+
+More details can be found here:
+https://hackmd.io/XhnYHKKuR6-LChhobvlT-g?view
+
+For now, the code for these two crates are in separate modules of this crate.
+The modules have the same name for simplicity. We also have a third module,
+`rustc_internal` which will expose APIs and definitions that allow users to
+gather information from internal MIR constructs that haven't been exposed in
+the `stable_mir` module.
diff --git a/compiler/rustc_smir/rust-toolchain.toml b/compiler/rustc_smir/rust-toolchain.toml
@@ -1,3 +1,3 @@
 [toolchain]
-channel = "nightly-2022-06-01"
+channel = "nightly-2023-02-28"
 components = [ "rustfmt", "rustc-dev" ]
diff --git a/compiler/rustc_smir/src/lib.rs b/compiler/rustc_smir/src/lib.rs
@@ -11,9 +11,9 @@
     test(attr(allow(unused_variables), deny(warnings)))
 )]
 #![cfg_attr(not(feature = "default"), feature(rustc_private))]
-#![deny(rustc::untranslatable_diagnostic)]
-#![deny(rustc::diagnostic_outside_of_impl)]
 
-pub mod mir;
+pub mod rustc_internal;
+pub mod stable_mir;
 
-pub mod very_unstable;
+// Make this module private for now since external users should not call these directly.
+mod rustc_smir;
diff --git a/compiler/rustc_smir/src/mir.rs b/compiler/rustc_smir/src/mir.rs
diff --git a/compiler/rustc_smir/src/rustc_internal/mod.rs b/compiler/rustc_smir/src/rustc_internal/mod.rs
@@ -0,0 +1,15 @@
+//! Module that implements the bridge between Stable MIR and internal compiler MIR.
+//!
+//! For that, we define APIs that will temporarily be public to 3P that exposes rustc internal APIs
+//! until stable MIR is complete.
+
+use crate::stable_mir;
+pub use rustc_span::def_id::{CrateNum, DefId};
+
+pub fn item_def_id(item: &stable_mir::CrateItem) -> DefId {
+    item.0
+}
+
+pub fn crate_num(item: &stable_mir::Crate) -> CrateNum {
+    item.id.into()
+}
diff --git a/compiler/rustc_smir/src/rustc_smir/mod.rs b/compiler/rustc_smir/src/rustc_smir/mod.rs
@@ -0,0 +1,48 @@
+//! Module that implements what will become the rustc side of Stable MIR.
+//!
+//! This module is responsible for building Stable MIR components from internal components.
+//!
+//! This module is not intended to be invoked directly by users. It will eventually
+//! become the public API of rustc that will be invoked by the `stable_mir` crate.
+//!
+//! For now, we are developing everything inside `rustc`, thus, we keep this module private.
+
+use crate::stable_mir::{self};
+use rustc_middle::ty::{tls::with, TyCtxt};
+use rustc_span::def_id::{CrateNum, LOCAL_CRATE};
+use tracing::debug;
+
+/// Get information about the local crate.
+pub fn local_crate() -> stable_mir::Crate {
+    with(|tcx| smir_crate(tcx, LOCAL_CRATE))
+}
+
+/// Retrieve a list of all external crates.
+pub fn external_crates() -> Vec<stable_mir::Crate> {
+    with(|tcx| tcx.crates(()).iter().map(|crate_num| smir_crate(tcx, *crate_num)).collect())
+}
+
+/// Find a crate with the given name.
+pub fn find_crate(name: &str) -> Option<stable_mir::Crate> {
+    with(|tcx| {
+        [LOCAL_CRATE].iter().chain(tcx.crates(()).iter()).find_map(|crate_num| {
+            let crate_name = tcx.crate_name(*crate_num).to_string();
+            (name == crate_name).then(|| smir_crate(tcx, *crate_num))
+        })
+    })
+}
+
+/// Retrieve all items of the local crate that have a MIR associated with them.
+pub fn all_local_items() -> stable_mir::CrateItems {
+    with(|tcx| {
+        tcx.mir_keys(()).iter().map(|item| stable_mir::CrateItem(item.to_def_id())).collect()
+    })
+}
+
+/// Build a stable mir crate from a given crate number.
+fn smir_crate(tcx: TyCtxt<'_>, crate_num: CrateNum) -> stable_mir::Crate {
+    let crate_name = tcx.crate_name(crate_num).to_string();
+    let is_local = crate_num == LOCAL_CRATE;
+    debug!(?crate_name, ?crate_num, "smir_crate");
+    stable_mir::Crate { id: crate_num.into(), name: crate_name, is_local }
+}
diff --git a/compiler/rustc_smir/src/stable_mir/mod.rs b/compiler/rustc_smir/src/stable_mir/mod.rs
@@ -0,0 +1,60 @@
+//! Module that implements the public interface to the Stable MIR.
+//!
+//! This module shall contain all type definitions and APIs that we expect 3P tools to invoke to
+//! interact with the compiler.
+//!
+//! The goal is to eventually move this module to its own crate which shall be published on
+//! [crates.io](https://crates.io).
+//!
+//! ## Note:
+//!
+//! There shouldn't be any direct references to internal compiler constructs in this module.
+//! If you need an internal construct, consider using `rustc_internal` or `rustc_smir`.
+
+use crate::rustc_internal;
+
+/// Use String for now but we should replace it.
+pub type Symbol = String;
+
+/// The number that identifies a crate.
+pub type CrateNum = usize;
+
+/// A unique identification number for each item accessible for the current compilation unit.
+pub type DefId = usize;
+
+/// A list of crate items.
+pub type CrateItems = Vec<CrateItem>;
+
+/// Holds information about a crate.
+#[derive(Clone, PartialEq, Eq, Debug)]
+pub struct Crate {
+    pub(crate) id: CrateNum,
+    pub name: Symbol,
+    pub is_local: bool,
+}
+
+/// Holds information about an item in the crate.
+/// For now, it only stores the item DefId. Use functions inside `rustc_internal` module to
+/// use this item.
+#[derive(Clone, PartialEq, Eq, Debug)]
+pub struct CrateItem(pub(crate) rustc_internal::DefId);
+
+/// Access to the local crate.
+pub fn local_crate() -> Crate {
+    crate::rustc_smir::local_crate()
+}
+
+/// Try to find a crate with the given name.
+pub fn find_crate(name: &str) -> Option<Crate> {
+    crate::rustc_smir::find_crate(name)
+}
+
+/// Try to find a crate with the given name.
+pub fn external_crates() -> Vec<Crate> {
+    crate::rustc_smir::external_crates()
+}
+
+/// Retrieve all items in the local crate that have a MIR associated with them.
+pub fn all_local_items() -> CrateItems {
+    crate::rustc_smir::all_local_items()
+}
diff --git a/compiler/rustc_smir/src/very_unstable.rs b/compiler/rustc_smir/src/very_unstable.rs
diff --git a/tests/ui-fulldeps/stable-mir/crate-info.rs b/tests/ui-fulldeps/stable-mir/crate-info.rs
@@ -0,0 +1,104 @@
+// run-pass
+// Test that users are able to use stable mir APIs to retrieve information of the current crate
+
+// ignore-stage-1
+// ignore-cross-compile
+// ignore-remote
+
+#![feature(rustc_private)]
+
+extern crate rustc_driver;
+extern crate rustc_hir;
+extern crate rustc_interface;
+extern crate rustc_middle;
+extern crate rustc_smir;
+
+use rustc_driver::{Callbacks, Compilation, RunCompiler};
+use rustc_hir::def::DefKind;
+use rustc_interface::{interface, Queries};
+use rustc_middle::ty::TyCtxt;
+use rustc_smir::{rustc_internal, stable_mir};
+use std::io::Write;
+
+const CRATE_NAME: &str = "input";
+
+/// This function uses the Stable MIR APIs to get information about the test crate.
+fn test_stable_mir(tcx: TyCtxt<'_>) {
+    // Get the local crate using stable_mir API.
+    let local = stable_mir::local_crate();
+    assert_eq!(&local.name, CRATE_NAME);
+
+    // Find items in the local crate.
+    let items = stable_mir::all_local_items();
+    assert!(has_item(tcx, &items, (DefKind::Fn, "foo_bar")));
+    assert!(has_item(tcx, &items, (DefKind::Fn, "foo::bar")));
+
+    // Find the `std` crate.
+    assert!(stable_mir::find_crate("std").is_some());
+}
+
+// Use internal API to find a function in a crate.
+fn has_item(tcx: TyCtxt, items: &stable_mir::CrateItems, item: (DefKind, &str)) -> bool {
+    items.iter().any(|crate_item| {
+        let def_id = rustc_internal::item_def_id(crate_item);
+        tcx.def_kind(def_id) == item.0 && tcx.def_path_str(def_id) == item.1
+    })
+}
+
+/// This test will generate and analyze a dummy crate using the stable mir.
+/// For that, it will first write the dummy crate into a file.
+/// It will invoke the compiler using a custom Callback implementation, which will
+/// invoke Stable MIR APIs after the compiler has finished its analysis.
+fn main() {
+    let path = "input.rs";
+    generate_input(&path).unwrap();
+    let args = vec![
+        "rustc".to_string(),
+        "--crate-type=lib".to_string(),
+        "--crate-name".to_string(),
+        CRATE_NAME.to_string(),
+        path.to_string(),
+    ];
+    rustc_driver::catch_fatal_errors(|| {
+        RunCompiler::new(&args, &mut SMirCalls {}).run().unwrap();
+    })
+    .unwrap();
+}
+
+struct SMirCalls {}
+
+impl Callbacks for SMirCalls {
+    /// Called after analysis. Return value instructs the compiler whether to
+    /// continue the compilation afterwards (defaults to `Compilation::Continue`)
+    fn after_analysis<'tcx>(
+        &mut self,
+        _compiler: &interface::Compiler,
+        queries: &'tcx Queries<'tcx>,
+    ) -> Compilation {
+        queries.global_ctxt().unwrap().enter(|tcx| {
+            test_stable_mir(tcx);
+        });
+        // No need to keep going.
+        Compilation::Stop
+    }
+}
+
+fn generate_input(path: &str) -> std::io::Result<()> {
+    let mut file = std::fs::File::create(path)?;
+    write!(
+        file,
+        r#"
+    mod foo {{
+        pub fn bar(i: i32) -> i64 {{
+            i as i64
+        }}
+    }}
+
+    pub fn foo_bar(x: i32, y: i32) -> i64 {{
+        let x_64 = foo::bar(x);
+        let y_64 = foo::bar(y);
+        x_64.wrapping_add(y_64)
+    }}"#
+    )?;
+    Ok(())
+}
