More questions about the FreePascal Language Reference (version 3.2.0), part 4

34. Are macpas LEAVE and CYCLE statements undocumented ?

35. Are macpas ellipsis (...) parameters undocumented ?

36. Is the macpas RETURN statement undocumented ?

37. Shouldn't the macpas "mwpascal" modifier be added to ?

call-modifiers = "register" | "cdecl" | "pascal" | "stdcall" | "safecall" | "inline" | "mwpascal" . call-modifiers = "cdecl" | "inline" | "local" | "nostackframe" | "overload" | "pascal" | "register" | "safecall" | "saveregisters" | "softfloat" | "stdcall" | "varargs" | "mwpascal".

38. Is a <constref-parameter> rule missing ?

constref-parameter = "CONSTREF" identifier-list [ ":" [ "ARRAY" "OF" ] type-identifier ] .

where

parameter-declaration = value-parameter | variable-parameter | out-parameter | constant-parameter | out-parameter | constref-parameter .

39. Section 16.7 defines a rule <library> and a rule <exports-clause>

library = library-header ";" [ uses-clause ] block "." . exports-clause = "exports" exports-list ";" .

The exports-clause rule however doesn't seem to be referenced anywhere in the syntax, so it may have to be added to the <library> rule ?

40. < recordoperator-definition> = <record-operator-definition> ?

41. Is it correct that <operator-definition> is referenced only in the record <component-list> rule ? I would expect something like an <operator-declaration-part> in various declaration rules.

(to be continued)

Adriaan van Os

Yes.

35. Are macpas ellipsis (...) parameters undocumented ?

Yes.

36. Is the macpas RETURN statement undocumented ?

Yes.

Yes. There are even more.

Yes.

To the library rule, but also to the unit rule, because it can also appear in a unit...

40. < recordoperator-definition> = <record-operator-definition> ?

Yes.

41. Is it correct that <operator-definition> is referenced only in the record <component-list> rule ? I would expect something like an <operator-declaration-part> in various declaration rules.

To the best of my knowledge, only records support operator definitions.

The 'operator' chapter handles 'global' operators which are at the level of global functions/procedures.

Michael.

And, according to Appendix D.6 of the Programmer's manual:

7. The cdecl modifier keyword can be abbreviated to C.

in macpas mode.

Regards,

Adriaan van Os

Well, according to Appendix D.6 of the Programmer's Manual, CYCLE is equivalent to CONTINUE and LEAVE to BREAK. Contrary to the information presented there, the CYCLE and LEAVE statements have been implemented for macpas mode.

I don't see BREAK and CONTINUE in the syntax diagrams either. I put together some examples (stressing that I am not a proponent of these C constructs).

procedure test;
begin
   while true do leave;
   writeln( 'leave makes true=false');
   while true do break;
   writeln( 'break makes true=false');
   writeln( 'cycle forever');
   while true do begin cycle; leave end;
   writeln( 'continue forever');
   while true do begin continue; break end;
end;

> 36. Is the macpas RETURN statement undocumented ?

In MetroWerks Pascal, we have

return-statement = "RETURN" [ expression ] .

In a procedure the [ expression ] part is skipped, in a function it assigns the function-result and then jumps to the end of the function. In a procedure, it just jumps to the end of the procedure. FreePascal allows RETURN in macpas mode to be used in a function only, so the following compiles only in a function.

  repeat return until false;
   writeln( 'the point of no return');

Ceterum censeo C esse delendam.

Regards,

Adriaan van Os

Added to the docs.

Michael.

That is because "break" and "continue" as well as leave are not keywords: you can use them as identifiers.

As such they are considered normal procedure calls and are documented in the system unit.

That explains why they're not in the syntax diagrams: when they appear, they are treated as procedure calls.

Michael.