[clang] [clang][CodeView] Prevent the input name from appearing in LF_BUILDINFO (PR #194140)

[via cfe-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-clang-codegen

@llvm/pr-subscribers-clang

Author: Alexandre Ganea (aganea)

<details>
<summary>Changes</summary>

The implicit contract of an `LF_BUILDINFO` record (represented in LLVM by [`BuildInfoRecord`](https://github.com/llvm/llvm-project/blob/6f0b55ec55f3e5e1ccc0d6b0d04a307479218768/llvm/include/llvm/DebugInfo/CodeView/TypeRecord.h#L667)) is that its `CommandLine` field should not contain the input source file — a separate `SourceFile` field is reserved for that.

When the command-line flattening was moved from `llvm/` to `clang/` in #<!-- -->106369, the comparison value used to identify and strip the source positional was switched from `MainSourceFile->getFilename()` (the full input path resolved by clang) to `CodeGenOpts.MainFileName` (just the basename, set via `-main-file-name`). As a result, when the driver is invoked with an absolute source path the cc1 positional is that absolute path and no longer matches `MainFileName`, so the source filename leaks into `CommandLine` as a trailing positional cc1 argument.

This is a regression in Clang 20. It breaks downstream tooling such as Live++, whose unity-splitting feature relies on the embedded command line being the cc1 invocation minus the source. Reported in #<!-- -->193900.

This PR restores the previous behavior by passing the resolved frontend input path(s) to `flattenClangCommandLine` and including them in the equality check that strips the source positional. The basename match against `MainFileName` is kept for the relative-input case. A regression test (and a symmetric relative-path test) is added to `clang/test/DebugInfo/Generic/codeview-buildinfo.c`.

Should fix #<!-- -->193900.

---
Full diff: https://github.com/llvm/llvm-project/pull/194140.diff


2 Files Affected:

- (modified) clang/lib/CodeGen/BackendUtil.cpp (+19-3) 
- (modified) clang/test/DebugInfo/Generic/codeview-buildinfo.c (+23) 


``````````diff
diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp
index f34c5c007f115..c47c7baf4d4b0 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -323,7 +323,8 @@ static bool actionRequiresCodeGen(BackendAction Action) {
 }
 
 static std::string flattenClangCommandLine(ArrayRef<std::string> Args,
-                                           StringRef MainFilename) {
+                                           StringRef MainFilename,
+                                           ArrayRef<StringRef> InputFiles) {
   if (Args.empty())
     return std::string{};
 
@@ -342,7 +343,15 @@ static std::string flattenClangCommandLine(ArrayRef<std::string> Args,
       i++; // Skip this argument and next one.
       continue;
     }
-    if (Arg.starts_with("-object-file-name") || Arg == MainFilename)
+    if (Arg.starts_with("-object-file-name"))
+      continue;
+    // Strip the source positional, matching either MainFilename (the
+    // -main-file-name basename) or one of the resolved frontend input paths
+    // (which is what the cc1 positional looks like for an absolute driver
+    // input). Avoid a generic basename match: it would also strip values of
+    // args like `-include <path>` whose trailing component happens to equal
+    // the source basename.
+    if (Arg == MainFilename || llvm::is_contained(InputFiles, Arg))
       continue;
     // Skip fmessage-length for reproducibility.
     if (Arg.starts_with("-fmessage-length"))
@@ -519,8 +528,15 @@ static bool initTargetOptions(const CompilerInstance &CI,
       Options.MCOptions.IASSearchPaths.push_back(
           Entry.IgnoreSysRoot ? Entry.Path : HSOpts.Sysroot + Entry.Path);
   Options.MCOptions.Argv0 = CodeGenOpts.Argv0 ? CodeGenOpts.Argv0 : "";
+  // Pass the resolved frontend inputs so flattenClangCommandLine can strip
+  // the cc1 source positional even when the driver received an absolute path
+  // (which won't match CodeGenOpts.MainFileName, that's just the basename).
+  SmallVector<StringRef, 1> InputFiles;
+  for (const auto &Input : CI.getFrontendOpts().Inputs)
+    if (Input.isFile())
+      InputFiles.push_back(Input.getFile());
   Options.MCOptions.CommandlineArgs = flattenClangCommandLine(
-      CodeGenOpts.CommandLineArgs, CodeGenOpts.MainFileName);
+      CodeGenOpts.CommandLineArgs, CodeGenOpts.MainFileName, InputFiles);
   Options.MCOptions.AsSecureLogFile = CodeGenOpts.AsSecureLogFile;
   Options.MCOptions.PPCUseFullRegisterNames =
       CodeGenOpts.PPCUseFullRegisterNames;
diff --git a/clang/test/DebugInfo/Generic/codeview-buildinfo.c b/clang/test/DebugInfo/Generic/codeview-buildinfo.c
index ee6f772944d84..01723cfae44c2 100644
--- a/clang/test/DebugInfo/Generic/codeview-buildinfo.c
+++ b/clang/test/DebugInfo/Generic/codeview-buildinfo.c
@@ -12,6 +12,18 @@
 // RUN: %clang_cl -gcodeview-command-line --target=i686-windows-msvc -Xclang -fmessage-length=100 /c /Z7 /Fo%t.obj -- %s
 // RUN: llvm-pdbutil dump --types %t.obj | FileCheck %s --check-prefix MESSAGELEN
 
+// The source filename must be stripped from the embedded cc1 command line
+// whether it's passed to the driver as an absolute or a relative path.
+// See https://github.com/llvm/llvm-project/issues/193900.
+
+// RUN: %clang_cl --target=i686-windows-msvc /c /Z7 /Fo%t.obj -- %s
+// RUN: llvm-pdbutil dump --types %t.obj | FileCheck %s --check-prefix ABSPATH
+
+// RUN: rm -rf %t.relpath && mkdir %t.relpath
+// RUN: cp %s %t.relpath/hello.cpp
+// RUN: cd %t.relpath && %clang_cl --target=i686-windows-msvc /c /Z7 /Fo:hello.obj -- hello.cpp
+// RUN: llvm-pdbutil dump --types %t.relpath/hello.obj | FileCheck %s --check-prefix RELPATH
+
 int main(void) { return 42; }
 
 // CHECK:                       Types (.debug$T)
@@ -45,3 +57,14 @@ int main(void) { return 42; }
 // MESSAGELEN: ============================================================
 // MESSAGELEN: 0x{{.+}} | LF_BUILDINFO [size = {{.+}}]
 // MESSAGELEN-NOT: -fmessage-length
+
+// The cmdline LF_STRING_ID quotes each cc1 arg, so a leaked source
+// positional appears as a quoted token. The FILEPATH LF_STRING_ID prints
+// the path without surrounding quotes, so it won't trigger the negative.
+// ABSPATH:     0x{{.+}} | LF_STRING_ID [size = {{.+}}] ID: <no type>, String: "-cc1
+// ABSPATH-NOT: "{{[^"]*[\\/]codeview-buildinfo\.c}}"
+// ABSPATH:     0x{{.+}} | LF_BUILDINFO [size = {{.+}}]
+
+// RELPATH:     0x{{.+}} | LF_STRING_ID [size = {{.+}}] ID: <no type>, String: "-cc1
+// RELPATH-NOT: "hello.cpp"
+// RELPATH:     0x{{.+}} | LF_BUILDINFO [size = {{.+}}]

``````````

</details>


https://github.com/llvm/llvm-project/pull/194140