[clang] [clang][modules-driver] Add scanner to detect C++20 module presence (PR #145220)

[Michael Spencer via cfe-commits]

================
@@ -4403,6 +4411,174 @@ void Driver::handleArguments(Compilation &C, DerivedArgList &Args,
   }
 }
 
+static void skipWhitespace(const char *&Ptr) {
+  while (isWhitespace(*Ptr))
+    ++Ptr;
+}
+
+// Returns the length of EOL, either 0 (no end-of-line), 1 (\n) or 2 (\r\n).
+static unsigned isEOL(const char *Ptr) {
+  if (*Ptr == '\0')
+    return 0;
+  if (*(Ptr + 1) != '\0' && isVerticalWhitespace(Ptr[0]) &&
+      isVerticalWhitespace(Ptr[1]) && Ptr[0] != Ptr[1])
+    return 2;
+  return !!isVerticalWhitespace(Ptr[0]);
+}
+
+static void skipLine(const char *&Ptr) {
+  for (;;) {
+    char LastNonWhitespace = ' ';
+    while (!isVerticalWhitespace(*Ptr) && *Ptr != '\0') {
+      if (!isHorizontalWhitespace(*Ptr))
+        LastNonWhitespace = *Ptr;
+      ++Ptr;
+    }
+
+    const unsigned Len = isEOL(Ptr);
+    if (!Len)
+      return;
+
+    Ptr += Len;
+    if (LastNonWhitespace != '\\')
+      break;
+  }
+}
+
+// Returns the length of a line splice sequence (including trailing
+// whitespace), or 0 if no line splice is found.
+static unsigned isLineSplice(const char *Start) {
+  if (*Start != '\\')
+    return 0;
+
+  const char *Ptr = Start + 1;
+  while (isHorizontalWhitespace(*Ptr))
+    ++Ptr;
+
+  if (unsigned Len = isEOL(Ptr))
+    return Ptr - Start + Len;
+  return 0;
+}
+
+static bool trySkipLineSplice(const char *&Ptr) {
+  if (unsigned Len = isLineSplice(Ptr); Len) {
+    Ptr += Len;
+    return true;
+  }
+  return false;
+}
+
+static bool trySkipDirective(const char *&Ptr) {
+  if (*Ptr != '#')
+    return false;
+
+  ++Ptr;
+  skipLine(Ptr);
+  return true;
+}
+
+static bool trySkipLineComment(const char *&Ptr) {
+  if (Ptr[0] != '/' || Ptr[1] != '/')
+    return false;
+
+  Ptr += 2;
+  skipLine(Ptr);
+  return true;
+}
+
+static bool trySkipBlockComment(const char *&Ptr) {
+  if (Ptr[0] != '/' || Ptr[1] != '*')
+    return false;
+
+  Ptr += 2;
+  while (*Ptr != '\0') {
+    if (Ptr[0] == '*' && Ptr[1] == '/') {
+      Ptr += 2; // '*/'
+      return true;
+    }
+    ++Ptr;
+  }
+  return true;
+}
+
+static bool trySkipComment(const char *&Ptr) {
+  return trySkipLineComment(Ptr) || trySkipBlockComment(Ptr);
+}
+
+// Skipps over comments and (non-module) directives
+static void skipToRelevantCXXModuleText(const char *&Ptr) {
+  while (*Ptr != '\0') {
+    skipWhitespace(Ptr);
+    if (trySkipComment(Ptr) || trySkipDirective(Ptr) || trySkipLineSplice(Ptr))
+      continue;
+    break; // Found relevant text!
+  }
+}
+
+static bool scanBufferForCXXModuleUsage(const llvm::MemoryBuffer &Buffer) {
+  const char *Ptr = Buffer.getBufferStart();
+  skipToRelevantCXXModuleText(Ptr);
+
+  // Check if buffer has enough bytes left to check for the module-related
+  // declaration fragment we want to check without making potentially
----------------
Bigcheese wrote:

```suggestion
  // declaration fragment we want to check without making the potentially
```

https://github.com/llvm/llvm-project/pull/145220