[clang-format] Improve detection of parameter declarations in K&R C

[owenpan]

Summary

Clean up the detection of parameter declarations in K&R C function definitions. Also make it more precise by requiring the second token after the r_paren to be either a star or keyword/identifier.

[owenpan]

Simplified the code slightly.

[MyDeveloperDay]

clang/lib/Format/UnwrappedLineParser.cpp

1025

should we have test cases showing these examples?

1029

Can you add unit tests for these cases?

1397

Can't next be null once again?

[owenpan]

clang/lib/Format/UnwrappedLineParser.cpp

1025

Probably not unless we are to include a test case for every keyword. I can add them if you insist.

1029

I think the existing test cases already cover this, which is a subset of the previous cases (the second token can be anything but tok::l_paren and tok::semi)?

1397

No because FormatTok is not tok::eof. If AllTokens[Position] is null, it would be an internal error, which is asserted at the top of isC78ParameterDecl().

[MyDeveloperDay]

LGTM

[MyDeveloperDay]

@owenpan Can we push all these commits to 13 RC2 via https://bugs.llvm.org/show_bug.cgi?id=51470

We need to mark the commits we want to cherry pick I think.

[owenpan]

In D108094#2951973, @MyDeveloperDay wrote:

@owenpan Can we push all these commits to 13 RC2 via https://bugs.llvm.org/show_bug.cgi?id=51470

We need to mark the commits we want to cherry pick I think.

Thanks! Updated 51470. Let me know if there is anything else for me to do.

clang/lib/Format/UnwrappedLineParser.cpp

 
 // This function checks whether a token starts the first parameter declaration
 // in a K&R C (aka C78) function definition, e.g.:
 //   int f(a, b)
 //   short a, b;
 //   {
 //      return a + b;
 //   }
-static bool isC78ParameterDecl(const FormatToken *Tok) {
-  if (!Tok)
+static bool isC78ParameterDecl(const FormatToken *Tok, const FormatToken *Next,
+                               const FormatToken *FuncName) {
+  assert(Tok);
+  assert(Next);
+  assert(FuncName);
+
+  if (FuncName->isNot(tok::identifier))
+    return false;
+
+  const FormatToken *Prev = FuncName->Previous;
+  if (!Prev || (Prev->isNot(tok::star) && !isC78Type(*Prev)))
     return false;
 
   if (!isC78Type(*Tok) &&
       !Tok->isOneOf(tok::kw_register, tok::kw_struct, tok::kw_union))
     return false;
 
+  if (Next->isNot(tok::star) && !Next->Tok.getIdentifierInfo())
+    return false;
+
   Tok = Tok->Previous;
   if (!Tok || Tok->isNot(tok::r_paren))
     return false;
 
   Tok = Tok->Previous;
   if (!Tok || Tok->isNot(tok::identifier))
     return false;
 
case tok::semi:
       return;
     case tok::r_brace:
       addUnwrappedLine();
       return;
     case tok::l_paren: {
       parseParens();
       // Break the unwrapped line if a K&R C function definition has a parameter
       // declaration.
-      if (!IsTopLevel || !Style.isCpp())
-        break;
-      if (!Previous || Previous->isNot(tok::identifier))
-        break;
-      const FormatToken *PrevPrev = Previous->Previous;
-      if (!PrevPrev || (!isC78Type(*PrevPrev) && PrevPrev->isNot(tok::star)))
+      if (!IsTopLevel || !Style.isCpp() || !Previous || FormatTok->is(tok::eof))
         break;
       const unsigned Position = Tokens->getPosition() + 1;
-      if (Position == AllTokens.size())
-        break;
       assert(Position < AllTokens.size());
-      const FormatToken *Next = AllTokens[Position];
-      if (Next && Next->isOneOf(tok::l_paren, tok::semi))
-        break;
-      if (isC78ParameterDecl(FormatTok)) {
+      if (isC78ParameterDecl(FormatTok, AllTokens[Position], Previous)) {
         addUnwrappedLine();
         return;
       }
       break;
     }
     case tok::kw_operator:
       nextToken();
       if (FormatTok->isBinaryOperator())