Re: [PATCH] Objective-C language support.

[Michael Snyder <msnyder@redhat.com>]

Andrew Cagney wrote:
> 
> > Adam Fedor wrote:
> >
> >>
> >> This patch adds Objective-C language support to gdb based upon a patch
> >> provided by Apple Computer Inc from their version of gdb. Note that the
> >> patch only contains changes to existing files. New files (objc-lang.h,
> >> objc-lang.c, objc-exp.y) and a gdb.objc testsuite directory are located at
> >>
> >> ftp://ftp.gnustep.org/pub/gnustep/contrib/gdb-objc-patch.tar.gz
> >
> >
> > Adam, this bit can't hurt anything, since it just adds some new enum
> > values.  Only one suggestion -- how about putting OP_THIS and OP_SELF
> > consecutively, because they're so similar?  Also please hold off on
> > the export of evaluate_subexp -- let's handle that separately.
> >
> > Other than that, approved.  Wait a couple days for objections, then
> > you can commit it.
> >
> >
> >
> >>         * expression.h: New ops NSSTRING, SELECTOR, MSGCALL, and SELF.
> >
> >
> >> Index: gdb/expression.h
> >> ===================================================================
> >> RCS file: /cvs/src/src/gdb/expression.h,v
> >> retrieving revision 1.4
> >> diff -u -p -r1.4 expression.h
> >> --- gdb/expression.h    1 Aug 2002 17:18:32 -0000       1.4
> >> +++ gdb/expression.h    17 Sep 2002 19:30:05 -0000
> >> @@ -181,6 +181,12 @@ enum exp_opcode
> >>         making three exp_elements.  */
> >>      OP_FUNCALL,
> >>
> >> +    /* OP_MSGCALL is followed by a string in the next exp_element and then an
> >> +       integer.  The string is the selector string.  The integer is the number
> >> +       of arguments to the message call.  That many plus one values are used,
> >> +       the first one being the object pointer.  This is an Objective C message */
> >> +    OP_MSGCALL,
> >> +
> 
> FYI, I think these should be renamed to OP_OBJC_xxxx so that it is very
> clear that these have Objective C semantics.  Otherwize I think it runs
> the risk of people trying to re-use these operators in other languages.
> 
> This would be consistent with the Ada OPS where I recommended
> OP_ADA_xxxx for the Ada specific extensions to the operator table.
> 
> Andrew

OK with me.

 
> >>      /* This is EXACTLY like OP_FUNCALL but is semantically different.
> >>         In F77, array subscript expressions, substring expressions
> >>         and function calls are  all exactly the same syntactically. They may
> >> @@ -273,11 +279,17 @@ enum exp_opcode
> >>      STRUCTOP_STRUCT,
> >>      STRUCTOP_PTR,
> >>
> >> -    /* C++ */
> >> -    /* OP_THIS is just a placeholder for the class instance variable.
> >> +    /* C++: OP_THIS is just a placeholder for the class instance variable.
> >>         It just comes in a tight (OP_THIS, OP_THIS) pair.  */
> >>      OP_THIS,
> >>
> >> +    /* Objective C: "@selector" pseudo-operator */
> >> +    OP_SELECTOR,
> >> +
> >> +    /* Objective-C: OP_SELF is just a placeholder for the class instance
> >> +       variable.  It just comes in a tight (OP_SELF, OP_SELF) pair.  */
> >> +    OP_SELF,
> >> +
> >>      /* OP_SCOPE surrounds a type name and a field name.  The type
> >>         name is encoded as one element, but the field name stays as
> >>         a string, which, of course, is variable length.  */
> >> @@ -305,7 +317,10 @@ enum exp_opcode
> >>      OP_NAME,
> >>
> >>      /* An unparsed expression.  Used for Scheme (for now at least) */
> >> -    OP_EXPRSTRING
> >> +    OP_EXPRSTRING,
> >> +
> >> +    /* An Objective C Foundation Class NSString constant */
> >> +    OP_NSSTRING,
> >>    };
> >>
> >>  union exp_element
> >> @@ -369,6 +384,10 @@ enum noside
> >>
> >>  extern struct value *evaluate_subexp_standard
> >>    (struct type *, struct expression *, int *, enum noside);
> >> +
> >> +extern struct value *evaluate_subexp (struct type *, struct expression *,
> >> +                                     int *, enum noside);
> >> +
> >>
> >>  /* From expprint.c */
> >>
> >
> >