Auto merge of #118518 - weihanglo:trim-paths-macos, r=<try>

fix(trim-paths): trim `SO` and `DW_AT_comp_dir` symbols for root DI node

This is one way to fix <#117652>.

## The issue

When `--remap-path-scope=object` is specified, user expect that there is
no local path embedded in final executables. Under `object` scope, the
current implementation only remap debug symbols if debug info is
splitted into its own file. In other words, when
`split-debuginfo=packed|unpacked` is set, rustc assumes there is no
embedded path in the final executable needing to be remapped.

However, this doesn't work.

* On Linux, the root `DW_AT_comp_dir` of a compile unit seems to go into the binary executables.
* On macOS, `SO` symbols are embedded in binary executables and libraries regardless a split-debuginfo file is built. Each `SO` symbol contains a path to the root source file of a debug info compile unit.

## The approach

Path of working directory in the root DI node seems to be embedded in
executables. Hence, we trim them when the scope of `unsplit-debuginfo`
is present, as if it is kinda a "unsplit" debuginfo.

## Unresolved issue

* Not sure where to add more test to consolidate it.
* Haven't investigate if we should apply the same logic to cranelift [here](https://github.com/rust-lang/rust/blob/64d7e0d0b61c460fbc882ae37c0f236756dd9c39/compiler/rustc_codegen_cranelift/src/debuginfo/mod.rs#L68-L80).
* Not sure if there is any other consequence doing this, but AFAIK debugging still works on macOS and Linux  with `profile.dev.trim-paths="object"` fairly (with some extra tweaks in cargo).

compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs

@@ -851,9 +851,22 @@ pub fn build_compile_unit_di_node<'ll, 'tcx>(
     // FIXME(#41252) Remove "clang LLVM" if we can get GDB and LLVM to play nice.
     let producer = format!("clang LLVM ({rustc_producer})");
 
-    use rustc_session::RemapFileNameExt;
     let name_in_debuginfo = name_in_debuginfo.to_string_lossy();
-    let work_dir = tcx.sess.opts.working_dir.for_codegen(tcx.sess).to_string_lossy();
+    debug!(?name_in_debuginfo, "build_compile_unit_di_node");
+
+    // Path of working directory in the root DI node seems to be embedded in
+    // executables. Hence, we trim them when the scope of `unsplit-debuginfo`
+    // is present, as if it is kinda a "unsplit" debuginfo.
+    use rustc_session::config::RemapPathScopeComponents;
+    use rustc_session::RemapFileNameExt;
+    let work_dir = tcx
+        .sess
+        .opts
+        .working_dir
+        .for_scope(tcx.sess, RemapPathScopeComponents::UNSPLIT_DEBUGINFO)
+        .to_string_lossy();
+    debug!(?work_dir, "build_compile_unit_di_node");
+
     let flags = "\0";
     let output_filenames = tcx.output_filenames(());
     let split_name = if tcx.sess.target_can_use_split_dwarf() {

tests/run-make/split-debuginfo/Makefile

@@ -1,6 +1,8 @@
 # ignore-cross-compile
 include ../tools.mk
 
+export HERE := $(shell pwd)
+
 all: off packed unpacked
 
 ifeq ($(UNAME),Darwin)
@@ -11,23 +13,77 @@ off:
 	[ ! -d $(TMPDIR)/foo.dSYM ]
 
 # Packed by default, but only if debuginfo is requested
-packed:
-	rm -rf $(TMPDIR)/*.dSYM
-	$(RUSTC) foo.rs
-	[ ! -d $(TMPDIR)/foo.dSYM ]
-	rm -rf $(TMPDIR)/*.dSYM
-	$(RUSTC) foo.rs -g
+packed: packed-remapped-scope packed-remapped-wrong-scope
+
+# - Debuginfo in binary file
+# - `.o` deleted
+# - `.dSYM` present
+# - in binary, paths from `N_SO` (source files) and `N_OSO` (object files) shouldn be remapped
+packed-remapped-scope:
+	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=packed -C debuginfo=2 \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		-Z remap-path-scope=object foo.rs -g
+	ls $(TMPDIR)/*.o && exit 1 || exit 0
 	[ -d $(TMPDIR)/foo.dSYM ]
-	rm -rf $(TMPDIR)/*.dSYM
-	$(RUSTC) foo.rs -g -C split-debuginfo=packed
+	# As of 2023-12, `OSO` should be the only thing that cannot be trimmed. See rust-lang/rust#116948
+	dsymutil -s $(TMPDIR)/foo | grep $(TMPDIR) || exit 1 # expected: `grep $(TMPDIR)` to exit 1
+	dsymutil -s $(TMPDIR)/foo | (! grep $(HERE)) || exit 1
+	rm -rf $(TMPDIR)/foo.dSYM
+	rm $(TMPDIR)/$(call BIN,foo)
+
+# - Debuginfo in binary file
+# - `.o` deleted
+# - `.dSYM` present
+# - in binary, paths from `N_SO` (source files) and `N_OSO` (object files) shouldn't be remapped
+packed-remapped-wrong-scope:
+	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=packed -C debuginfo=2 \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		-Z remap-path-scope=macro foo.rs -g
+	ls $(TMPDIR)/*.o && exit 1 || exit 0
 	[ -d $(TMPDIR)/foo.dSYM ]
-	rm -rf $(TMPDIR)/*.dSYM
+	dsymutil -s $(TMPDIR)/foo | grep 'N_OSO' | grep $(TMPDIR) || exit 1
+	dsymutil -s $(TMPDIR)/foo | grep 'N_SO' | grep $(HERE) || exit 1
+	rm -rf $(TMPDIR)/foo.dSYM
+	rm $(TMPDIR)/$(call BIN,foo)
 
 # Object files are preserved with unpacked and `dsymutil` isn't run
-unpacked:
-	$(RUSTC) foo.rs -g -C split-debuginfo=unpacked
+unpacked: unpacked-remapped-scope unpacked-remapped-wrong-scope
+
+# - Debuginfo in object files
+# - `.o` present
+# - `.dSYM` never created
+# - in binary, paths from `N_SO` (source files) and `N_OSO` (object files) should be remapped
+unpacked-remapped-scope:
+	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=unpacked -C debuginfo=2 \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		-Z remap-path-scope=object foo.rs -g
+	ls $(TMPDIR)/*.o
+	[ ! -d $(TMPDIR)/foo.dSYM ]
+	# As of 2023-12, `OSO` should be the only thing that cannot be trimmed. See rust-lang/rust#116948
+	dsymutil -s $(TMPDIR)/foo | grep $(TMPDIR) || exit 1 # expected: `grep $(TMPDIR)` to exit 1
+	dsymutil -s $(TMPDIR)/foo | (! grep $(HERE)) || exit 1
+	rm $(TMPDIR)/*.o
+	rm $(TMPDIR)/$(call BIN,foo)
+
+# - Debuginfo in object files
+# - `.o` present,
+# - `.dSYM` never created
+# - in binary, paths from `N_SO` (source files) and `N_OSO` (object files) shouldn't be remapped
+unpacked-remapped-wrong-scope:
+	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=unpacked -C debuginfo=2 \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		-Z remap-path-scope=macro foo.rs -g
 	ls $(TMPDIR)/*.o
 	[ ! -d $(TMPDIR)/foo.dSYM ]
+	dsymutil -s $(TMPDIR)/foo | grep 'N_OSO' | (grep $(TMPDIR)) || exit 1
+	dsymutil -s $(TMPDIR)/foo | grep 'N_SO' | (grep $(HERE)) || exit 1
+	rm $(TMPDIR)/*.o
+	rm $(TMPDIR)/$(call BIN,foo)
+
 else
 ifdef IS_WINDOWS
 # Windows only supports packed debuginfo - nothing to test.
@@ -113,8 +169,12 @@ packed-remapped: packed-remapped-split packed-remapped-single packed-remapped-sc
 # - `.dwp` present
 packed-remapped-split:
 	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=packed -C debuginfo=2 \
-		-Z split-dwarf-kind=split --remap-path-prefix $(TMPDIR)=/a foo.rs -g
-	objdump -Wi $(TMPDIR)/foo | grep DW_AT_GNU_dwo_name | (! grep $(TMPDIR)) || exit 1
+		-Z split-dwarf-kind=split \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		foo.rs -g
+	readelf -wi $(TMPDIR)/foo | (! grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | (! grep $(HERE)) || exit 1
 	ls $(TMPDIR)/*.o && exit 1 || exit 0
 	ls $(TMPDIR)/*.dwo && exit 1 || exit 0
 	rm $(TMPDIR)/foo.dwp
@@ -127,8 +187,12 @@ packed-remapped-split:
 # - `.dwp` present
 packed-remapped-single:
 	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=packed -C debuginfo=2 \
-		-Z split-dwarf-kind=single --remap-path-prefix $(TMPDIR)=/a foo.rs -g
-	objdump -Wi $(TMPDIR)/foo | grep DW_AT_GNU_dwo_name | (! grep $(TMPDIR)) || exit 1
+		-Z split-dwarf-kind=single \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		foo.rs -g
+	readelf -wi $(TMPDIR)/foo | (! grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | (! grep $(HERE)) || exit 1
 	ls $(TMPDIR)/*.o && exit 1 || exit 0
 	ls $(TMPDIR)/*.dwo && exit 1 || exit 0
 	rm $(TMPDIR)/foo.dwp
@@ -141,9 +205,12 @@ packed-remapped-single:
 # - `.dwp` present
 packed-remapped-scope:
 	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=packed -C debuginfo=2 \
-		-Z split-dwarf-kind=single --remap-path-prefix $(TMPDIR)=/a \
-		-Z remap-path-scope=split-debuginfo-path foo.rs -g
-	objdump -Wi $(TMPDIR)/foo | grep DW_AT_GNU_dwo_name | (! grep $(TMPDIR)) || exit 1
+		-Z split-dwarf-kind=single \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		-Z remap-path-scope=object foo.rs -g
+	readelf -wi $(TMPDIR)/foo | (! grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | (! grep $(HERE)) || exit 1
 	ls $(TMPDIR)/*.o && exit 1 || exit 0
 	ls $(TMPDIR)/*.dwo && exit 1 || exit 0
 	rm $(TMPDIR)/foo.dwp
@@ -156,9 +223,12 @@ packed-remapped-scope:
 # - `.dwp` present
 packed-remapped-wrong-scope:
 	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=packed -C debuginfo=2 \
-		-Z split-dwarf-kind=single --remap-path-prefix $(TMPDIR)=/a \
+		-Z split-dwarf-kind=single \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(TMPDIR)=/b \
 		-Z remap-path-scope=macro foo.rs -g
-	objdump -Wi $(TMPDIR)/foo | grep DW_AT_GNU_dwo_name | (grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | (grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | grep $(HERE) || exit 1
 	ls $(TMPDIR)/*.o && exit 1 || exit 0
 	ls $(TMPDIR)/*.dwo && exit 1 || exit 0
 	rm $(TMPDIR)/foo.dwp
@@ -269,8 +339,12 @@ unpacked-remapped: unpacked-remapped-split unpacked-remapped-single unpacked-rem
 # - `.dwp` never created
 unpacked-remapped-split:
 	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=unpacked -C debuginfo=2 \
-		-Z split-dwarf-kind=split --remap-path-prefix $(TMPDIR)=/a foo.rs -g
-	objdump -Wi $(TMPDIR)/foo | grep DW_AT_GNU_dwo_name | (! grep $(TMPDIR)) || exit 1
+		-Z split-dwarf-kind=split \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		foo.rs -g
+	readelf -wi $(TMPDIR)/foo | (! grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | (! grep $(HERE)) || exit 1
 	ls $(TMPDIR)/*.o && exit 1 || exit 0
 	rm $(TMPDIR)/*.dwo
 	ls $(TMPDIR)/*.dwp && exit 1 || exit 0
@@ -283,8 +357,12 @@ unpacked-remapped-split:
 # - `.dwp` never created
 unpacked-remapped-single:
 	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=unpacked -C debuginfo=2 \
-		-Z split-dwarf-kind=single --remap-path-prefix $(TMPDIR)=/a foo.rs -g
-	objdump -Wi $(TMPDIR)/foo | grep DW_AT_GNU_dwo_name | (! grep $(TMPDIR)) || exit 1
+		-Z split-dwarf-kind=single \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		foo.rs -g
+	readelf -wi $(TMPDIR)/foo | (! grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | (! grep $(HERE)) || exit 1
 	rm $(TMPDIR)/*.o
 	ls $(TMPDIR)/*.dwo && exit 1 || exit 0
 	ls $(TMPDIR)/*.dwp && exit 1 || exit 0
@@ -297,9 +375,12 @@ unpacked-remapped-single:
 # - `.dwp` never created
 unpacked-remapped-scope:
 	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=unpacked -C debuginfo=2 \
-		-Z split-dwarf-kind=single --remap-path-prefix $(TMPDIR)=/a \
-		-Z remap-path-scope=split-debuginfo-path foo.rs -g
-	objdump -Wi $(TMPDIR)/foo | grep DW_AT_GNU_dwo_name | (! grep $(TMPDIR)) || exit 1
+		-Z split-dwarf-kind=single \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
+		-Z remap-path-scope=object foo.rs -g
+	readelf -wi $(TMPDIR)/foo | (! grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | (! grep $(HERE)) || exit 1
 	rm $(TMPDIR)/*.o
 	ls $(TMPDIR)/*.dwo && exit 1 || exit 0
 	ls $(TMPDIR)/*.dwp && exit 1 || exit 0
@@ -312,9 +393,12 @@ unpacked-remapped-scope:
 # - `.dwp` never created
 unpacked-remapped-wrong-scope:
 	$(RUSTC) $(UNSTABLEOPTS) -C split-debuginfo=unpacked -C debuginfo=2 \
-		-Z split-dwarf-kind=single --remap-path-prefix $(TMPDIR)=/a \
+		-Z split-dwarf-kind=single \
+		--remap-path-prefix $(TMPDIR)=/a \
+		--remap-path-prefix $(HERE)=/b \
 		-Z remap-path-scope=macro foo.rs -g
-	objdump -Wi $(TMPDIR)/foo | grep DW_AT_GNU_dwo_name | (grep $(TMPDIR)) || exit 1
+	readelf -wi $(TMPDIR)/foo | grep $(TMPDIR) || exit 1
+	readelf -wi $(TMPDIR)/foo | grep $(HERE) || exit 1
 	rm $(TMPDIR)/*.o
 	ls $(TMPDIR)/*.dwo && exit 1 || exit 0
 	ls $(TMPDIR)/*.dwp && exit 1 || exit 0