[clang] Patch series to reapply #118734 and substantially improve it (PR #120534)

[Chandler Carruth via cfe-commits]

================
@@ -68,23 +70,144 @@ enum ID {
   FirstTSBuiltin
 };
 
+// The info used to represent each builtin.
 struct Info {
-  llvm::StringLiteral Name;
-  const char *Type, *Attributes;
-  const char *Features;
+  // Rather than store pointers to the string literals describing these four
+  // aspects of builtins, we store offsets into a common string table.
+  struct StrOffsets {
+    llvm::StringTable::Offset Name;
+    llvm::StringTable::Offset Type;
+    llvm::StringTable::Offset Attributes;
+    llvm::StringTable::Offset Features;
+  } Offsets;
+
   HeaderDesc Header;
   LanguageID Langs;
 };
 
+// A constexpr function to construct an infos array from X-macros.
+//
+// The input array uses the same data structure, but the offsets are actually
+// _lengths_ when input. This is all we can compute from the X-macro approach to
+// builtins. This function will convert these lengths into actual offsets to a
+// string table built up through sequentially appending strings with the given
+// lengths.
+template <size_t N>
+static constexpr std::array<Info, N> MakeInfos(std::array<Info, N> Infos) {
+  // Translate lengths to offsets. We start past the initial empty string at
+  // offset zero.
+  unsigned Offset = 1;
+  for (auto &I : Infos) {
+    Info::StrOffsets NewOffsets = {};
+    NewOffsets.Name = Offset;
+    Offset += I.Offsets.Name.value();
+    NewOffsets.Type = Offset;
+    Offset += I.Offsets.Type.value();
+    NewOffsets.Attributes = Offset;
+    Offset += I.Offsets.Attributes.value();
+    NewOffsets.Features = Offset;
+    Offset += I.Offsets.Features.value();
+    I.Offsets = NewOffsets;
+  }
+  return Infos;
+}
+
+// A detail macro used below to emit a string literal that, after string literal
+// concatenation, ends up triggering the `-Woverlength-strings` warning. While
+// the warning is useful in general to catch accidentally excessive strings,
+// here we are creating them intentionally.
+//
+// This relies on a subtle aspect of `_Pragma`: that the *diagnostic* ones don't
+// turn into actual tokens that would disrupt string literal concatenation.
+#ifdef __clang__
+#define CLANG_BUILTIN_DETAIL_STR_TABLE(S)                                      \
+  _Pragma("clang diagnostic push")                                             \
+      _Pragma("clang diagnostic ignored \"-Woverlength-strings\"")             \
----------------
chandlerc wrote:

Sadly, yes. I don't understand why either. =[

https://github.com/llvm/llvm-project/pull/120534