[llvm-ar] Use COFF archive format for COFF targets. (#82642)

Detect COFF files by default and allow specifying it with --format
argument.

This is important for ARM64EC, which uses a separated symbol map for EC
symbols. Since K_COFF is mostly compatible with K_GNU, this shouldn't
really make a difference for other targets.

5 files changed, 114 insertions, 15 deletions

diff --git a/llvm/include/llvm/Object/Archive.h b/llvm/include/llvm/Object/Archive.h
index 3dd99a4..66f0793 100644
--- a/llvm/include/llvm/Object/Archive.h
+++ b/llvm/include/llvm/Object/Archive.h
@@ -339,6 +339,7 @@ public:
   Kind kind() const { return (Kind)Format; }
   bool isThin() const { return IsThin; }
   static object::Archive::Kind getDefaultKindForHost();
+  static object::Archive::Kind getDefaultKindForTriple(Triple &T);
 
   child_iterator child_begin(Error &Err, bool SkipInternal = true) const;
   child_iterator child_end() const;
diff --git a/llvm/lib/Object/Archive.cpp b/llvm/lib/Object/Archive.cpp
index e447e5b..d3fdcd9 100644
--- a/llvm/lib/Object/Archive.cpp
+++ b/llvm/lib/Object/Archive.cpp
@@ -969,12 +969,19 @@ Archive::Archive(MemoryBufferRef Source, Error &Err)
   Err = Error::success();
 }
 
+object::Archive::Kind Archive::getDefaultKindForTriple(Triple &T) {
+  if (T.isOSDarwin())
+    return object::Archive::K_DARWIN;
+  if (T.isOSAIX())
+    return object::Archive::K_AIXBIG;
+  if (T.isOSWindows())
+    return object::Archive::K_COFF;
+  return object::Archive::K_GNU;
+}
+
 object::Archive::Kind Archive::getDefaultKindForHost() {
   Triple HostTriple(sys::getProcessTriple());
-  return HostTriple.isOSDarwin()
-             ? object::Archive::K_DARWIN
-             : (HostTriple.isOSAIX() ? object::Archive::K_AIXBIG
-                                     : object::Archive::K_GNU);
+  return getDefaultKindForTriple(HostTriple);
 }
 
 Archive::child_iterator Archive::child_begin(Error &Err,
diff --git a/llvm/lib/Object/ArchiveWriter.cpp b/llvm/lib/Object/ArchiveWriter.cpp
index 155926a..02f7252 100644
--- a/llvm/lib/Object/ArchiveWriter.cpp
+++ b/llvm/lib/Object/ArchiveWriter.cpp
@@ -62,12 +62,16 @@ object::Archive::Kind NewArchiveMember::detectKindFromObject() const {
   Expected<std::unique_ptr<object::ObjectFile>> OptionalObject =
       object::ObjectFile::createObjectFile(MemBufferRef);
 
-  if (OptionalObject)
-    return isa<object::MachOObjectFile>(**OptionalObject)
-               ? object::Archive::K_DARWIN
-               : (isa<object::XCOFFObjectFile>(**OptionalObject)
-                      ? object::Archive::K_AIXBIG
-                      : object::Archive::K_GNU);
+  if (OptionalObject) {
+    if (isa<object::MachOObjectFile>(**OptionalObject))
+      return object::Archive::K_DARWIN;
+    if (isa<object::XCOFFObjectFile>(**OptionalObject))
+      return object::Archive::K_AIXBIG;
+    if (isa<object::COFFObjectFile>(**OptionalObject) ||
+        isa<object::COFFImportFile>(**OptionalObject))
+      return object::Archive::K_COFF;
+    return object::Archive::K_GNU;
+  }
 
   // Squelch the error in case we had a non-object file.
   consumeError(OptionalObject.takeError());
@@ -80,10 +84,7 @@ object::Archive::Kind NewArchiveMember::detectKindFromObject() const {
             MemBufferRef, file_magic::bitcode, &Context)) {
       auto &IRObject = cast<object::IRObjectFile>(**ObjOrErr);
       auto TargetTriple = Triple(IRObject.getTargetTriple());
-      return TargetTriple.isOSDarwin()
-                 ? object::Archive::K_DARWIN
-                 : (TargetTriple.isOSAIX() ? object::Archive::K_AIXBIG
-                                           : object::Archive::K_GNU);
+      return object::Archive::getDefaultKindForTriple(TargetTriple);
     } else {
       // Squelch the error in case this was not a SymbolicFile.
       consumeError(ObjOrErr.takeError());
diff --git a/llvm/test/tools/llvm-ar/coff-symtab.test b/llvm/test/tools/llvm-ar/coff-symtab.test
new file mode 100644
index 0000000..50d08fb
--- /dev/null
+++ b/llvm/test/tools/llvm-ar/coff-symtab.test
@@ -0,0 +1,85 @@
+Verify that llvm-ar uses COFF archive format by ensuring that archive map is sorted.
+
+RUN: rm -rf %t.dif && split-file %s %t.dir && cd %t.dir
+
+RUN: yaml2obj coff-symtab.yaml -o coff-symtab.obj
+RUN: llvm-ar crs out.a coff-symtab.obj
+RUN: llvm-nm --print-armap out.a | FileCheck %s
+
+RUN: llvm-as coff-symtab.ll -o coff-symtab.bc
+RUN: llvm-ar crs out2.a coff-symtab.bc
+RUN: llvm-nm --print-armap out2.a | FileCheck %s
+
+RUN: yaml2obj elf.yaml -o coff-symtab.o
+RUN: llvm-ar crs --format coff out3.a coff-symtab.o
+RUN: llvm-nm --print-armap out3.a | FileCheck %s
+
+CHECK: Archive map
+CHECK-NEXT: a in coff-symtab
+CHECK-NEXT: b in coff-symtab
+CHECK-NEXT: c in coff-symtab
+CHECK-EMPTY:
+
+#--- coff-symtab.yaml
+--- !COFF
+header:
+  Machine:           IMAGE_FILE_MACHINE_UNKNOWN
+  Characteristics:   [  ]
+sections:
+  - Name:            .text
+    Characteristics: [ IMAGE_SCN_CNT_CODE, IMAGE_SCN_MEM_EXECUTE, IMAGE_SCN_MEM_READ ]
+    Alignment:       4
+    SectionData:     ''
+symbols:
+  - Name:            b
+    Value:           0
+    SectionNumber:   1
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_FUNCTION
+    StorageClass:    IMAGE_SYM_CLASS_EXTERNAL
+  - Name:            c
+    Value:           0
+    SectionNumber:   1
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_FUNCTION
+    StorageClass:    IMAGE_SYM_CLASS_EXTERNAL
+  - Name:            a
+    Value:           0
+    SectionNumber:   1
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_FUNCTION
+    StorageClass:    IMAGE_SYM_CLASS_EXTERNAL
+...
+
+
+#--- coff-symtab.ll
+target triple = "x86_64-unknown-windows-msvc"
+
+define void @b() { ret void }
+define void @c() { ret void }
+define void @a() { ret void }
+
+#--- elf.yaml
+--- !ELF
+FileHeader:
+  Class:             ELFCLASS64
+  Data  :            ELFDATA2LSB
+  Type:              ET_REL
+  Machine:           EM_X86_64
+Sections:
+  - Name:            .text
+    Type:            SHT_PROGBITS
+    Flags:           [ SHF_ALLOC, SHF_EXECINSTR ]
+    AddressAlign:    0x0000000000000004
+    Content:         ''
+Symbols:
+  - Name:            b
+    Binding:         STB_GLOBAL
+    Section:         .text
+  - Name:            c
+    Binding:         STB_GLOBAL
+    Section:         .text
+  - Name:            a
+    Binding:         STB_GLOBAL
+    Section:         .text
+...
diff --git a/llvm/tools/llvm-ar/llvm-ar.cpp b/llvm/tools/llvm-ar/llvm-ar.cpp
index c880030..c58a85c 100644
--- a/llvm/tools/llvm-ar/llvm-ar.cpp
+++ b/llvm/tools/llvm-ar/llvm-ar.cpp
@@ -82,6 +82,7 @@ static void printArHelp(StringRef ToolName) {
     =darwin             -   darwin
     =bsd                -   bsd
     =bigarchive         -   big archive (AIX OS)
+    =coff               -   coff
   --plugin=<string>     - ignored for compatibility
   -h --help             - display this help and exit
   --output              - the directory to extract archive members to
@@ -193,7 +194,7 @@ static SmallVector<const char *, 256> PositionalArgs;
 static bool MRI;
 
 namespace {
-enum Format { Default, GNU, BSD, DARWIN, BIGARCHIVE, Unknown };
+enum Format { Default, GNU, COFF, BSD, DARWIN, BIGARCHIVE, Unknown };
 }
 
 static Format FormatType = Default;
@@ -1044,6 +1045,9 @@ static void performWriteOperation(ArchiveOperation Operation,
   case GNU:
     Kind = object::Archive::K_GNU;
     break;
+  case COFF:
+    Kind = object::Archive::K_COFF;
+    break;
   case BSD:
     if (Thin)
       fail("only the gnu format has a thin mode");
@@ -1376,6 +1380,7 @@ static int ar_main(int argc, char **argv) {
                        .Case("darwin", DARWIN)
                        .Case("bsd", BSD)
                        .Case("bigarchive", BIGARCHIVE)
+                       .Case("coff", COFF)
                        .Default(Unknown);
       if (FormatType == Unknown)
         fail(std::string("Invalid format ") + Match);