[lld] [lld][ELF] Add --debug-names to create merged .debug_names. (PR #86508)

[David Blaikie via llvm-commits]

================
@@ -2711,6 +2714,606 @@ static uint32_t computeGdbHash(StringRef s) {
   return h;
 }
 
+// 4-byte alignment ensures that values in the hash lookup table and the name
+// table are aligned.
+DebugNamesBaseSection::DebugNamesBaseSection()
+    : SyntheticSection(0, SHT_PROGBITS, 4, ".debug_names") {}
+
+// Get the size of the .debug_names section header in bytes for DWARF32:
+static uint32_t getDebugNamesHeaderSize(uint32_t augmentationStringSize) {
+  return /* unit length */ 4 +
+         /* version */ 2 +
+         /* padding */ 2 +
+         /* CU count */ 4 +
+         /* TU count */ 4 +
+         /* Foreign TU count */ 4 +
+         /* Bucket Count */ 4 +
+         /* Name Count */ 4 +
+         /* Abbrev table size */ 4 +
+         /* Augmentation string size */ 4 +
+         /* Augmentation string */ augmentationStringSize;
+}
+
+void DebugNamesBaseSection::parseDebugNames(
+    InputChunk &inputChunk, OutputChunk &chunk,
+    DWARFDataExtractor &namesExtractor, DataExtractor &strExtractor,
+    function_ref<SmallVector<uint32_t, 0>(
+        const DWARFDebugNames::Header &,
+        const DWARFDebugNames::DWARFDebugNamesOffsets &)>
+        readOffsets) {
+  const LLDDWARFSection namesSec = inputChunk.section;
+  DenseMap<uint32_t, IndexEntry *> offsetMap;
+  for (const DWARFDebugNames::NameIndex &ni : *inputChunk.llvmDebugNames) {
+    NameData &nd = inputChunk.nameData.emplace_back();
+    nd.hdr = ni.getHeader();
+    if (nd.hdr.Format != DwarfFormat::DWARF32) {
+      errorOrWarn(toString(namesSec.sec) + Twine(": unsupported DWARF64"));
----------------
dwblaikie wrote:

Grammatically, maybe this should be "DWARF64 is unsupported"? ("unsupported version: 5" makes sense grammatically, being "5 is an unsupported version" but "unsupported DWARF64" doesn't parse the same - the DWARF spec calls it the "64-bit DWARF format" so I guess the equivalent to "unsupported version: 5" would be "unsupported DWARF format: 64-bit"? but that's still a bit clunky & even if the spec doesn't say "DWARF64", it's how we usually refer to it, so not using that name might be more confusing than helpful)

Though admittedly, "DWARF64 is unsupported" sounds like a generic statement, not very explicit that "this file is using DWARF64, and that is unsupported" - though "unsupported version: 5" is pretty similarly problematic... 

https://github.com/llvm/llvm-project/pull/86508