[RFC] Support .comment section like GCC/Clang (!llvm.ident)

compiler/rustc_codegen_llvm/src/context.rs

@@ -33,6 +33,7 @@ use rustc_target::abi::{
 use rustc_target::spec::{HasTargetSpec, RelocModel, Target, TlsModel};
 use smallvec::SmallVec;
 
+use libc::c_uint;
 use std::cell::{Cell, RefCell};
 use std::ffi::CStr;
 use std::str;
@@ -349,6 +350,23 @@ pub unsafe fn create_module<'ll>(
         );
     }
 
+    // Insert `llvm.ident` metadata.
+    //
+    // On the wasm targets it will get hooked up to the "producer" sections
+    // `processed-by` information.
+    let rustc_producer =
+        format!("rustc version {}", option_env!("CFG_VERSION").expect("CFG_VERSION"));
+    let name_metadata = llvm::LLVMMDStringInContext(
+        llcx,
+        rustc_producer.as_ptr().cast(),
+        rustc_producer.as_bytes().len() as c_uint,
+    );
+    llvm::LLVMAddNamedMetadataOperand(
+        llmod,
+        cstr!("llvm.ident").as_ptr(),
+        llvm::LLVMMDNodeInContext(llcx, &name_metadata, 1),
+    );
+
     llmod
 }
 

compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs

@@ -888,21 +888,6 @@ pub fn build_compile_unit_di_node<'ll, 'tcx>(
             llvm::LLVMAddNamedMetadataOperand(debug_context.llmod, llvm_gcov_ident.as_ptr(), val);
         }
 
-        // Insert `llvm.ident` metadata on the wasm targets since that will
-        // get hooked up to the "producer" sections `processed-by` information.
-        if tcx.sess.target.is_like_wasm {
-            let name_metadata = llvm::LLVMMDStringInContext(
-                debug_context.llcontext,
-                rustc_producer.as_ptr().cast(),
-                rustc_producer.as_bytes().len() as c_uint,
-            );
-            llvm::LLVMAddNamedMetadataOperand(
-                debug_context.llmod,
-                cstr!("llvm.ident").as_ptr(),
-                llvm::LLVMMDNodeInContext(debug_context.llcontext, &name_metadata, 1),
-            );
-        }
-
         return unit_metadata;
     };
 

compiler/rustc_llvm/llvm-wrapper/PassWrapper.cpp

@@ -1366,6 +1366,11 @@ LLVMRustPrepareThinLTOImport(const LLVMRustThinLTOData *Data, LLVMModuleRef M,
     if (WasmCustomSections)
       WasmCustomSections->eraseFromParent();
 
+    // `llvm.ident` named metadata also gets duplicated.
+    auto *llvmIdent = (*MOrErr)->getNamedMetadata("llvm.ident");
+    if (llvmIdent)
+      llvmIdent->eraseFromParent();
+
     return MOrErr;
   };
   bool ClearDSOLocal = clearDSOLocalOnDeclarations(Mod, Target);

tests/codegen/llvm-ident.rs

@@ -0,0 +1,15 @@
+// Verifies that the `!llvm.ident` named metadata is emitted.
+//
+// revisions: NONE OPT DEBUG
+//
+// [OPT] compile-flags: -Copt-level=2
+// [DEBUG] compile-flags: -Cdebuginfo=2
+
+// The named metadata should contain a single metadata node (see
+// `LLVMRustPrepareThinLTOImport` for details).
+// CHECK: !llvm.ident = !{![[ID:[0-9]+]]}
+
+// In addition, check that the metadata node has the expected content.
+// CHECK: ![[ID]] = !{!"rustc version 1.{{.*}}"}
+
+fn main() {}

tests/run-make/comment-section/Makefile

@@ -0,0 +1,15 @@
+include ../tools.mk
+
+# only-linux
+
+all:
+	echo 'fn main(){}' | $(RUSTC) - --emit=link,obj -Csave-temps --target=$(TARGET)
+
+	# Check linked output has a `.comment` section with the expected content.
+	readelf -p '.comment' $(TMPDIR)/rust_out | $(CGREP) -F 'rustc version 1.'
+
+	# Check all object files (including temporary outputs) have a `.comment`
+	# section with the expected content.
+	set -e; for f in $(TMPDIR)/*.o; do \
+		readelf -p '.comment' $$f | $(CGREP) -F 'rustc version 1.'; \
+	done

tests/run-make/llvm-ident/Makefile

@@ -0,0 +1,19 @@
+include ../tools.mk
+
+# only-linux
+
+all:
+	# `-Ccodegen-units=16 -Copt-level=2` is used here to trigger thin LTO
+	# across codegen units to test deduplication of the named metadata
+	# (see `LLVMRustPrepareThinLTOImport` for details).
+	echo 'fn main(){}' | $(RUSTC) - --emit=link,obj -Csave-temps -Ccodegen-units=16 -Copt-level=2 --target=$(TARGET)
+
+	# `llvm-dis` is used here since `--emit=llvm-ir` does not emit LLVM IR
+	# for temporary outputs.
+	"$(LLVM_BIN_DIR)"/llvm-dis $(TMPDIR)/*.bc
+
+	# Check LLVM IR files (including temporary outputs) have `!llvm.ident`
+	# named metadata, reusing the related codegen test.
+	set -e; for f in $(TMPDIR)/*.ll; do \
+		$(LLVM_FILECHECK) --input-file $$f ../../codegen/llvm-ident.rs; \
+	done