[WIP] Support compiling rustc to WebAssembly

Author: bjorn3

.gitignore

@@ -79,3 +79,8 @@ package.json
 tests/rustdoc-gui/src/**.lock
 
 # Before adding new lines, see the comment at the top.
+
+/wasi-sdk-*
+/wabt-*
+/tmp
+/rust_out

Cargo.lock

@@ -217,6 +217,12 @@ dependencies = [
  "object 0.32.2",
 ]
 
+[[package]]
+name = "arbitrary"
+version = "1.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7d5a26814d8dcb93b0e5a0ff3c6d80a8843bafb21b39e8e18a6f05471870e110"
+
 [[package]]
 name = "arrayvec"
 version = "0.7.4"
@@ -834,6 +840,121 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "cranelift-bforest"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5b3775cc6cc00c90d29eebea55feedb2b0168e23f5415bab7859c4004d7323d1"
+dependencies = [
+ "cranelift-entity",
+]
+
+[[package]]
+name = "cranelift-codegen"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "637f3184ba5bfa48d425bad1d2e4faf5fcf619f5e0ca107edc6dc02f589d4d74"
+dependencies = [
+ "bumpalo",
+ "cranelift-bforest",
+ "cranelift-codegen-meta",
+ "cranelift-codegen-shared",
+ "cranelift-control",
+ "cranelift-entity",
+ "cranelift-isle",
+ "gimli",
+ "hashbrown 0.14.3",
+ "log",
+ "regalloc2",
+ "smallvec",
+ "target-lexicon",
+]
+
+[[package]]
+name = "cranelift-codegen-meta"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e4b35b8240462341d94d31aab807cad704683988708261aecae3d57db48b7212"
+dependencies = [
+ "cranelift-codegen-shared",
+]
+
+[[package]]
+name = "cranelift-codegen-shared"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8f3cd1555aa9df1d6d8375732de41b4cb0d787006948d55b6d004d521e9efeb0"
+
+[[package]]
+name = "cranelift-control"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "14b31a562a10e98ab148fa146801e20665c5f9eda4fce9b2c5a3836575887d74"
+dependencies = [
+ "arbitrary",
+]
+
+[[package]]
+name = "cranelift-entity"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "af1e0467700a3f4fccf5feddbaebdf8b0eb82535b06a9600c4bc5df40872e75d"
+
+[[package]]
+name = "cranelift-frontend"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6cb918ee2c23939262efd1b99d76a21212ac7bd35129582133e21a22a6ff0467"
+dependencies = [
+ "cranelift-codegen",
+ "log",
+ "smallvec",
+ "target-lexicon",
+]
+
+[[package]]
+name = "cranelift-isle"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "966e4cfb23cf6d7f1d285d53a912baaffc5f06bcd9c9b0a2d8c66a184fae534b"
+
+[[package]]
+name = "cranelift-module"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cec3c5879608f8073782974e2c6dda461eb4038c754e1ab02904eb94a0ee0e9f"
+dependencies = [
+ "anyhow",
+ "cranelift-codegen",
+ "cranelift-control",
+]
+
+[[package]]
+name = "cranelift-native"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bea803aadfc4aabdfae7c3870f1b1f6dd4332f4091859e9758ef5fca6bf8cc87"
+dependencies = [
+ "cranelift-codegen",
+ "libc",
+ "target-lexicon",
+]
+
+[[package]]
+name = "cranelift-object"
+version = "0.106.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "db2734474e6b011263cf0ea0ca36378a29e8a9795f8a71b46b3fbb1f7657173c"
+dependencies = [
+ "anyhow",
+ "cranelift-codegen",
+ "cranelift-control",
+ "cranelift-module",
+ "log",
+ "object 0.32.2",
+ "target-lexicon",
+]
+
 [[package]]
 name = "crc32fast"
 version = "1.4.0"
@@ -1639,6 +1760,15 @@ dependencies = [
  "thiserror",
 ]
 
+[[package]]
+name = "hashbrown"
+version = "0.13.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "43a3c133739dddd0d2990f9a4bdf8eb4b21ef50e4851ca85ab661199821d510e"
+dependencies = [
+ "ahash",
+]
+
 [[package]]
 name = "hashbrown"
 version = "0.14.3"
@@ -1976,7 +2106,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "168fb715dda47215e360912c096649d23d58bf392ac62f73919e831745e40f26"
 dependencies = [
  "equivalent",
- "hashbrown",
+ "hashbrown 0.14.3",
  "rustc-rayon",
  "serde",
 ]
@@ -2638,7 +2768,7 @@ dependencies = [
  "compiler_builtins",
  "crc32fast",
  "flate2",
- "hashbrown",
+ "hashbrown 0.14.3",
  "indexmap",
  "memchr",
  "rustc-std-workspace-alloc",
@@ -3215,6 +3345,19 @@ dependencies = [
  "thiserror",
 ]
 
+[[package]]
+name = "regalloc2"
+version = "0.9.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ad156d539c879b7a24a363a2016d77961786e71f48f2e2fc8302a92abd2429a6"
+dependencies = [
+ "hashbrown 0.13.2",
+ "log",
+ "rustc-hash",
+ "slice-group-by",
+ "smallvec",
+]
+
 [[package]]
 name = "regex"
 version = "1.8.4"
@@ -3646,6 +3789,37 @@ dependencies = [
  "tracing",
 ]
 
+[[package]]
+name = "rustc_codegen_cranelift"
+version = "0.1.0"
+dependencies = [
+ "cranelift-codegen",
+ "cranelift-frontend",
+ "cranelift-module",
+ "cranelift-native",
+ "cranelift-object",
+ "gimli",
+ "indexmap",
+ "jobserver",
+ "object 0.32.2",
+ "rustc_ast",
+ "rustc_codegen_ssa",
+ "rustc_data_structures",
+ "rustc_errors",
+ "rustc_fs_util",
+ "rustc_hir",
+ "rustc_incremental",
+ "rustc_index",
+ "rustc_metadata",
+ "rustc_middle",
+ "rustc_monomorphize",
+ "rustc_session",
+ "rustc_span",
+ "rustc_target",
+ "smallvec",
+ "target-lexicon",
+]
+
 [[package]]
 name = "rustc_codegen_llvm"
 version = "0.0.0"
@@ -4108,6 +4282,7 @@ dependencies = [
  "rustc_attr",
  "rustc_borrowck",
  "rustc_builtin_macros",
+ "rustc_codegen_cranelift",
  "rustc_codegen_llvm",
  "rustc_codegen_ssa",
  "rustc_const_eval",
@@ -5140,6 +5315,12 @@ dependencies = [
  "autocfg",
 ]
 
+[[package]]
+name = "slice-group-by"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "826167069c09b99d56f31e9ae5c99049e932a98c9dc2dac47645b08dbbf76ba7"
+
 [[package]]
 name = "smallvec"
 version = "1.13.2"
@@ -5245,7 +5426,7 @@ dependencies = [
  "core",
  "dlmalloc",
  "fortanix-sgx-abi",
- "hashbrown",
+ "hashbrown 0.14.3",
  "hermit-abi",
  "libc",
  "miniz_oxide",
@@ -5444,6 +5625,12 @@ dependencies = [
  "xattr",
 ]
 
+[[package]]
+name = "target-lexicon"
+version = "0.12.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e1fc403891a21bcfb7c37834ba66a547a8f402146eba7265b5a6d88059c9ff2f"
+
 [[package]]
 name = "tempfile"
 version = "3.10.1"
@@ -5552,7 +5739,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4db52ee8fec06e119b692ef3dd2c4cf621a99204c1b8c47407870ed050305b9b"
 dependencies = [
  "gimli",
- "hashbrown",
+ "hashbrown 0.14.3",
  "object 0.32.2",
  "tracing",
 ]

comment.txt

@@ -0,0 +1,18 @@
+Got a new version up. This time all logic to handle using different codegen backends for the different stages, so you don't need to do a multi stage process.
+
+* Make sure you are running on Linux.
+* Clone https://github.com/bjorn3/rust
+* Checkout the `compile_rustc_for_wasm10` branch
+* Download and extract https://github.com/WebAssembly/wasi-sdk/releases/download/wasi-sdk-20/wasi-sdk-20.0-linux.tar.gz into the rust checkout.
+* Run `./x.py install` to build the compiler and install it into the dist dir.
+* Use `wasmtime run --dir tmp::/ --dir dist dist/bin/rustc.wasm --sysroot dist --target x86_64-unknown-linux-gnu` to run rustc.
+
+Note that linking is not supported as wasi doesn't allow spawning executables. You can use something like the following to compile using the wasm32-wasi rustc and then link using native gcc:
+
+```
+$ mkdir tmp
+$ echo 'fn main() { println!("Hello World!"); }' | wasmtime run -Sthreads=y -Spreview2=n --dir tmp::/ --dir dist --env RUST_MIN_STACK=16777216 dist/bin/rustc.wasm - --sysroot dist --target x86_64-unknown-linux-gnu -Csave-temps
+$ gcc -fuse-ld=lld tmp/rmeta*/lib.rmeta tmp/rust_out.* dist/lib/rustlib/x86_64-unknown-linux-gnu/lib/lib*.rlib -o rust_out
+$ ./rust_out
+Hello World!
+```

compiler/rustc/Cargo.toml

@@ -28,6 +28,7 @@ features = ['unprefixed_malloc_on_supported_platforms']
 [features]
 # tidy-alphabetical-start
 jemalloc = ['jemalloc-sys']
+cranelift = ['rustc_driver_impl/cranelift']
 llvm = ['rustc_driver_impl/llvm']
 max_level_info = ['rustc_driver_impl/max_level_info']
 rustc_use_parallel_compiler = ['rustc_driver_impl/rustc_use_parallel_compiler']

compiler/rustc_codegen_cranelift/Cargo.toml

@@ -4,7 +4,7 @@ version = "0.1.0"
 edition = "2021"
 
 [lib]
-crate-type = ["dylib"]
+crate-type = ["lib", "dylib"]
 
 [dependencies]
 # These have to be in sync with each other
@@ -22,6 +22,22 @@ indexmap = "2.0.0"
 libloading = { version = "0.8.0", optional = true }
 smallvec = "1.8.1"
 
+jobserver = "0.1.22"
+rustc_middle = { path = "../rustc_middle" }
+rustc_ast = { path = "../rustc_ast" }
+rustc_codegen_ssa = { path = "../rustc_codegen_ssa" }
+rustc_data_structures = { path = "../rustc_data_structures" }
+rustc_errors = { path = "../rustc_errors" }
+rustc_fs_util = { path = "../rustc_fs_util" }
+rustc_hir = { path = "../rustc_hir" }
+rustc_incremental = { path = "../rustc_incremental" }
+rustc_index = { path = "../rustc_index" }
+rustc_metadata = { path = "../rustc_metadata" }
+rustc_monomorphize = { path = "../rustc_monomorphize" }
+rustc_session = { path = "../rustc_session" }
+rustc_span = { path = "../rustc_span" }
+rustc_target = { path = "../rustc_target" }
+
 [patch.crates-io]
 # Uncomment to use local checkout of cranelift
 #cranelift-codegen = { path = "../wasmtime/cranelift/codegen" }

compiler/rustc_codegen_cranelift/src/lib.rs

@@ -27,8 +27,8 @@ extern crate rustc_span;
 extern crate rustc_target;
 
 // This prevents duplicating functions and statics that are already part of the host rustc process.
-#[allow(unused_extern_crates)]
-extern crate rustc_driver;
+//#[allow(unused_extern_crates)]
+//extern crate rustc_driver;
 
 use std::any::Any;
 use std::cell::{Cell, RefCell};
@@ -350,7 +350,7 @@ fn build_isa(sess: &Session, backend_config: &BackendConfig) -> Arc<dyn isa::Tar
 }
 
 /// This is the entrypoint for a hot plugged rustc_codegen_cranelift
-#[no_mangle]
+//#[no_mangle]
 pub fn __rustc_codegen_backend() -> Box<dyn CodegenBackend> {
     Box::new(CraneliftCodegenBackend { config: RefCell::new(None) })
 }

compiler/rustc_codegen_ssa/src/back/link.rs

@@ -104,10 +104,17 @@ pub fn link_binary<'a>(
         });
 
         if outputs.outputs.should_link() {
-            let tmpdir = TempFileBuilder::new()
-                .prefix("rustc")
-                .tempdir()
-                .unwrap_or_else(|error| sess.dcx().emit_fatal(errors::CreateTempDir { error }));
+            let tmpdir = if cfg!(target_family = "wasm") {
+                TempFileBuilder::new()
+                    .prefix("rustc")
+                    .tempdir_in(std::env::current_dir().unwrap())
+                    .unwrap_or_else(|error| sess.dcx().emit_fatal(errors::CreateTempDir { error }))
+            } else {
+                TempFileBuilder::new()
+                    .prefix("rustc")
+                    .tempdir()
+                    .unwrap_or_else(|error| sess.dcx().emit_fatal(errors::CreateTempDir { error }))
+            };
             let path = MaybeTempDir::new(tmpdir, sess.opts.cg.save_temps);
             let output = out_filename(
                 sess,

compiler/rustc_codegen_ssa/src/back/linker.rs

@@ -126,7 +126,9 @@ pub fn get_linker<'a>(
             new_path.extend(env::split_paths(&path));
         }
     }
-    cmd.env("PATH", env::join_paths(new_path).unwrap());
+    if cfg!(not(target_family = "wasm")) {
+        cmd.env("PATH", env::join_paths(new_path).unwrap());
+    }
 
     // FIXME: Move `/LIBPATH` addition for uwp targets from the linker construction
     // to the linker args construction.