[LLD][COFF] Ensure .bss is merged at the end of a section. (#137677)

Because it is full of zeros, it is expected that as much of it as
possible is elided from the actual image, and that cannot happen if
there is initialized data in the section after it.

Author: jeremyd2019

lld/COFF/Writer.cpp

@@ -215,6 +215,7 @@ class Writer {
   void appendImportThunks();
   void locateImportTables();
   void createExportTable();
+  void mergeSection(const std::map<StringRef, StringRef>::value_type &p);
   void mergeSections();
   void sortECChunks();
   void appendECImportTables();
@@ -1566,6 +1567,30 @@ void Writer::createSymbolAndStringTable() {
   fileSize = alignTo(fileOff, ctx.config.fileAlign);
 }
 
+void Writer::mergeSection(const std::map<StringRef, StringRef>::value_type &p) {
+  StringRef toName = p.second;
+  if (p.first == toName)
+    return;
+  StringSet<> names;
+  while (true) {
+    if (!names.insert(toName).second)
+      Fatal(ctx) << "/merge: cycle found for section '" << p.first << "'";
+    auto i = ctx.config.merge.find(toName);
+    if (i == ctx.config.merge.end())
+      break;
+    toName = i->second;
+  }
+  OutputSection *from = findSection(p.first);
+  OutputSection *to = findSection(toName);
+  if (!from)
+    return;
+  if (!to) {
+    from->name = toName;
+    return;
+  }
+  to->merge(from);
+}
+
 void Writer::mergeSections() {
   llvm::TimeTraceScope timeScope("Merge sections");
   if (!pdataSec->chunks.empty()) {
@@ -1594,28 +1619,16 @@ void Writer::mergeSections() {
   }
 
   for (auto &p : ctx.config.merge) {
-    StringRef toName = p.second;
-    if (p.first == toName)
-      continue;
-    StringSet<> names;
-    while (true) {
-      if (!names.insert(toName).second)
-        Fatal(ctx) << "/merge: cycle found for section '" << p.first << "'";
-      auto i = ctx.config.merge.find(toName);
-      if (i == ctx.config.merge.end())
-        break;
-      toName = i->second;
-    }
-    OutputSection *from = findSection(p.first);
-    OutputSection *to = findSection(toName);
-    if (!from)
-      continue;
-    if (!to) {
-      from->name = toName;
-      continue;
-    }
-    to->merge(from);
+    if (p.first != ".bss")
+      mergeSection(p);
   }
+
+  // Because .bss contains all zeros, it should be merged at the end of
+  // whatever section it is being merged into (usually .data) so that the image
+  // need not actually contain all of the zeros.
+  auto it = ctx.config.merge.find(".bss");
+  if (it != ctx.config.merge.end())
+    mergeSection(*it);
 }
 
 // EC targets may have chunks of various architectures mixed together at this

lld/test/COFF/merge-data-bss.test

@@ -0,0 +1,92 @@
+# RUN: yaml2obj %s -o %t.obj
+# RUN: lld-link /out:%t.exe /entry:main /subsystem:console /force \
+# RUN:   /merge:.other=.data %t.obj /debug
+# RUN: llvm-readobj --sections %t.exe | FileCheck %s
+
+# CHECK:      Name: .data
+# CHECK-NEXT: VirtualSize: 0x2018
+# CHECK-NEXT: VirtualAddress: 0x3000
+# CHECK-NEXT: RawDataSize: 512
+# CHECK-NOT: Name: .other
+
+--- !COFF
+header:
+  Machine:         IMAGE_FILE_MACHINE_AMD64
+  Characteristics: [  ]
+sections:
+  - Name:            .text
+    Characteristics: [ IMAGE_SCN_CNT_CODE, IMAGE_SCN_MEM_EXECUTE, IMAGE_SCN_MEM_READ ]
+    Alignment:       4
+    SectionData:     '90'
+    SizeOfRawData:   1
+  - Name:            .data
+    Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_READ, IMAGE_SCN_MEM_WRITE ]
+    Alignment:       4
+    SectionData:     '010000000000000002'
+    SizeOfRawData:   9
+  - Name:            .bss
+    Characteristics: [ IMAGE_SCN_CNT_UNINITIALIZED_DATA, IMAGE_SCN_MEM_READ, IMAGE_SCN_MEM_WRITE ]
+    Alignment:       4
+    SectionData:     ''
+    SizeOfRawData:   8192
+  - Name:            .other
+    Characteristics: [ IMAGE_SCN_CNT_INITIALIZED_DATA, IMAGE_SCN_MEM_READ, IMAGE_SCN_MEM_WRITE ]
+    Alignment:       4
+    SectionData:     '030000000000000004'
+    SizeOfRawData:   9
+symbols:
+  - Name:            .text
+    Value:           0
+    SectionNumber:   1
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_NULL
+    StorageClass:    IMAGE_SYM_CLASS_STATIC
+    SectionDefinition:
+      Length:          1
+      NumberOfRelocations: 0
+      NumberOfLinenumbers: 0
+      CheckSum:        4027552580
+      Number:          1
+  - Name:            .data
+    Value:           0
+    SectionNumber:   2
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_NULL
+    StorageClass:    IMAGE_SYM_CLASS_STATIC
+    SectionDefinition:
+      Length:          9
+      NumberOfRelocations: 0
+      NumberOfLinenumbers: 0
+      CheckSum:        4185224559
+      Number:          2
+  - Name:            .bss
+    Value:           0
+    SectionNumber:   3
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_NULL
+    StorageClass:    IMAGE_SYM_CLASS_STATIC
+    SectionDefinition:
+      Length:          8192
+      NumberOfRelocations: 0
+      NumberOfLinenumbers: 0
+      CheckSum:        0
+      Number:          3
+  - Name:            .other
+    Value:           0
+    SectionNumber:   4
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_NULL
+    StorageClass:    IMAGE_SYM_CLASS_STATIC
+    SectionDefinition:
+      Length:          9
+      NumberOfRelocations: 0
+      NumberOfLinenumbers: 0
+      CheckSum:        1054931164
+      Number:          4
+  - Name:            main
+    Value:           0
+    SectionNumber:   1
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_NULL
+    StorageClass:    IMAGE_SYM_CLASS_EXTERNAL
+...