Auto merge of rust-lang#135768 - jieyouxu:migrate-symbol-mangling-has…

…hed, r=<try>

tests: Port `symbol-mangling-hashed` to rmake.rs

Part of rust-lang#121876.

This PR supersedes rust-lang#128567 and is co-authored with `@lolbinarycat.`

### Summary

This PR ports `tests/run-make/symbol-mangling-hashed` to rmake.rs. Notable differences when compared to the Makefile version includes:

- It's no longer limited to linux + x86_64 only. In particular, this now is exercised on darwin and windows (esp. msvc) too.
- The test uses `object` crate to be more precise in the filtering, and avoids relying on parsing the human-readable `nm` output for *some* `nm` in the given environment (which isn't really a thing on msvc anyway, and `llvm-nm` doesn't handle msvc dylibs AFAICT).
- Dump the symbols satisfying various criteria on test failure to make it hopefully less of a pain to debug if it ever fails in CI.

### Review advice

- Best reviewed commit-by-commit.
- I'm not *super* sure about the msvc logic, would benefit from a MSVC (PE/COFF) expert taking a look.

---

try-job: x86_64-msvc
try-job: i686-msvc-1
try-job: i686-mingw
try-job: x86_64-mingw-1
try-job: x86_64-apple-1
try-job: aarch64-apple
try-job: test-various

src/tools/run-make-support/src/external_deps/rustc.rs

@@ -215,6 +215,18 @@ impl Rustc {
         self
     }
 
+    /// Specify option of `-C symbol-mangling-version`.
+    pub fn symbol_mangling_version(&mut self, option: &str) -> &mut Self {
+        self.cmd.arg(format!("-Csymbol-mangling-version={option}"));
+        self
+    }
+
+    /// Specify `-C prefer-dynamic`.
+    pub fn prefer_dynamic(&mut self) -> &mut Self {
+        self.cmd.arg(format!("-Cprefer-dynamic"));
+        self
+    }
+
     /// Specify error format to use
     pub fn error_format(&mut self, format: &str) -> &mut Self {
         self.cmd.arg(format!("--error-format={format}"));

src/tools/run-make-support/src/lib.rs

@@ -47,7 +47,9 @@ pub use wasmparser;
 // tidy-alphabetical-end
 
 // Re-exports of external dependencies.
-pub use external_deps::{c_build, c_cxx_compiler, clang, htmldocck, llvm, python, rustc, rustdoc};
+pub use external_deps::{
+    cargo, c_build, c_cxx_compiler, clang, htmldocck, llvm, python, rustc, rustdoc
+};
 
 // These rely on external dependencies.
 pub use c_cxx_compiler::{Cc, Gcc, cc, cxx, extra_c_flags, extra_cxx_flags, gcc};
@@ -79,7 +81,10 @@ pub use env::{env_var, env_var_os, set_current_dir};
 pub use run::{cmd, run, run_fail, run_with_args};
 
 /// Helpers for checking target information.
-pub use targets::{is_aix, is_darwin, is_msvc, is_windows, llvm_components_contain, target, uname, apple_os};
+pub use targets::{
+    apple_os, is_aix, is_darwin, is_msvc, is_windows, is_windows_gnu, llvm_components_contain,
+    target, uname,
+};
 
 /// Helpers for building names of output artifacts that are potentially target-specific.
 pub use artifact_names::{
@@ -104,4 +109,3 @@ pub use assertion_helpers::{
 pub use string::{
     count_regex_matches_in_files_with_extension, invalid_utf8_contains, invalid_utf8_not_contains,
 };
-use crate::external_deps::cargo;

src/tools/run-make-support/src/symbols.rs

@@ -2,18 +2,32 @@ use std::path::Path;
 
 use object::{self, Object, ObjectSymbol, SymbolIterator};
 
-/// Iterate through the symbols in an object file.
-///
-/// Uses a callback because `SymbolIterator` does not own its data.
+/// Given an [`object::File`], find the exported dynamic symbol names via
+/// [`object::Object::exports`]. This does **not** impose any filters on the specific dynamic
+/// symbols, e.g. if they are global or local, if they are defined or not, and in which section the
+/// dynamic symbols reside in.
+#[track_caller]
+pub fn exported_dynamic_symbol_names<'file>(file: &'file object::File<'file>) -> Vec<&'file str> {
+    file.exports()
+        .unwrap()
+        .into_iter()
+        .filter_map(|sym| std::str::from_utf8(sym.name()).ok())
+        .collect()
+}
+
+/// Iterate through the symbols in an object file. See [`object::Object::symbols`].
 ///
 /// Panics if `path` is not a valid object file readable by the current user.
+#[track_caller]
 pub fn with_symbol_iter<P, F, R>(path: P, func: F) -> R
 where
     P: AsRef<Path>,
     F: FnOnce(&mut SymbolIterator<'_, '_>) -> R,
 {
-    let raw_bytes = crate::fs::read(path);
-    let f = object::File::parse(raw_bytes.as_slice()).expect("unable to parse file");
+    let path = path.as_ref();
+    let blob = crate::fs::read(path);
+    let f = object::File::parse(&*blob)
+        .unwrap_or_else(|e| panic!("failed to parse `{}`: {e}", path.display()));
     let mut iter = f.symbols();
     func(&mut iter)
 }
@@ -24,6 +38,7 @@ where
 /// `path` contain a substring listed in `substrings`.
 ///
 /// Panics if `path` is not a valid object file readable by the current user.
+#[track_caller]
 pub fn any_symbol_contains(path: impl AsRef<Path>, substrings: &[&str]) -> bool {
     with_symbol_iter(path, |syms| {
         for sym in syms {

src/tools/run-make-support/src/targets.rs

@@ -22,6 +22,12 @@ pub fn is_msvc() -> bool {
     target().contains("msvc")
 }
 
+/// Check if target is windows-gnu.
+#[must_use]
+pub fn is_windows_gnu() -> bool {
+    target().ends_with("windows-gnu")
+}
+
 /// Check if target uses macOS.
 #[must_use]
 pub fn is_darwin() -> bool {

src/tools/tidy/src/allowed_run_make_makefiles.txt

@@ -1,5 +1,4 @@
 run-make/cat-and-grep-sanity-check/Makefile
 run-make/jobserver-error/Makefile
 run-make/split-debuginfo/Makefile
-run-make/symbol-mangling-hashed/Makefile
 run-make/translation/Makefile

tests/run-make/symbol-mangling-hashed/rmake.rs

@@ -0,0 +1,116 @@
+// ignore-tidy-linelength
+//! Basic smoke test for the unstable option `-C symbol_mangling_version=hashed` which aims to
+//! replace full symbol mangling names based on hash digests to shorten symbol name lengths in
+//! dylibs for space savings.
+//!
+//! # References
+//!
+//! - MCP #705: Provide option to shorten symbol names by replacing them with a digest:
+//!   <https://github.com/rust-lang/compiler-team/issues/705>.
+//! - Implementation PR: <https://github.com/rust-lang/rust/pull/118636>.
+//! - PE format: <https://learn.microsoft.com/en-us/windows/win32/debug/pe-format>.
+
+//@ ignore-cross-compile
+
+#![deny(warnings)]
+
+use run_make_support::symbols::exported_dynamic_symbol_names;
+use run_make_support::{bin_name, cwd, dynamic_lib_name, object, rfs, rustc};
+
+fn main() {
+    rustc()
+        .input("a_dylib.rs")
+        .prefer_dynamic()
+        .arg("-Zunstable-options")
+        .symbol_mangling_version("hashed")
+        .metadata("foo")
+        .run();
+
+    rustc()
+        .input("a_rlib.rs")
+        .prefer_dynamic()
+        .arg("-Zunstable-options")
+        .symbol_mangling_version("hashed")
+        .metadata("bar")
+        .run();
+
+    rustc().input("b_dylib.rs").library_search_path(cwd()).prefer_dynamic().run();
+    rustc().input("b_bin.rs").library_search_path(cwd()).prefer_dynamic().run();
+
+    // Check hashed symbol name
+
+    {
+        let dylib_filename = dynamic_lib_name("a_dylib");
+        println!("checking dylib `{dylib_filename}`");
+
+        let dylib_blob = rfs::read(&dylib_filename);
+        let dylib_file = object::File::parse(&*dylib_blob)
+            .unwrap_or_else(|e| panic!("failed to parse `{dylib_filename}`: {e}"));
+
+        let dynamic_symbols = exported_dynamic_symbol_names(&dylib_file);
+
+        if dynamic_symbols.iter().filter(|sym| sym.contains("hello")).count() != 0 {
+            eprintln!("exported dynamic symbols: {:#?}", dynamic_symbols);
+            panic!("expected no occurrence of `hello`");
+        }
+
+        if dynamic_symbols.iter().filter(|sym| sym.starts_with("_RNxC7a_dylib")).count() != 2 {
+            eprintln!("exported dynamic symbols: {:#?}", dynamic_symbols);
+            panic!("expected two dynamic symbols starting with `_RNxC7a_dylib`");
+        }
+    }
+
+    {
+        let so_filename = dynamic_lib_name("b_dylib");
+        println!("checking so `{so_filename}`");
+
+        let so_blob = rfs::read(&so_filename);
+        let so_file = object::File::parse(&*so_blob)
+            .unwrap_or_else(|e| panic!("failed to parse `{so_filename}`: {e}"));
+
+        let dynamic_symbols = exported_dynamic_symbol_names(&so_file);
+
+        if dynamic_symbols
+            .iter()
+            .filter(|sym| sym.contains("b_dylib") && sym.contains("hello"))
+            .count()
+            != 1
+        {
+            eprintln!("exported dynamic symbols: {:#?}", dynamic_symbols);
+            panic!("expected one occurrence of mangled `hello`");
+        }
+
+        if dynamic_symbols.iter().filter(|sym| sym.starts_with("_RNxC6a_rlib")).count() != 2 {
+            eprintln!("exported dynamic symbols: {:#?}", dynamic_symbols);
+            panic!("expected two exported symbols starting with `_RNxC6a_rlib`");
+        }
+
+        if dynamic_symbols.iter().any(|sym| sym.starts_with("_RNxC7a_dylib")) {
+            eprintln!("exported dynamic symbols: {:#?}", dynamic_symbols);
+            panic!("did not expect any symbols starting with `_RNxC7a_dylib`");
+        }
+    }
+
+    {
+        let bin_filename = bin_name("b_bin");
+        println!("checking bin `{bin_filename}`");
+
+        let bin_blob = rfs::read(&bin_filename);
+        let bin_file = object::File::parse(&*bin_blob)
+            .unwrap_or_else(|e| panic!("failed to parse `{bin_filename}`: {e}"));
+
+        let dynamic_symbols = exported_dynamic_symbol_names(&bin_file);
+
+        if dynamic_symbols.iter().any(|sym| {
+            sym.starts_with("_RNxC6a_rlib")
+                || sym.starts_with("_RNxC7a_dylib")
+                || (sym.contains("b_dylib") && sym.contains("hello"))
+        }) {
+            eprintln!("dynamic symbols: {:#?}", dynamic_symbols);
+            panic!(
+                "did not expect any symbols to (1) start with `_RNxC6a_rlib` or (2) start with \
+                     `_RNxC7a_dylib` or (3) to be of the form `*b_dylib*hello*`"
+            );
+        }
+    }
+}