Date: Sat, 25 Feb 1995 12:12:38 +0900
From: Stephen Turnbull <turnbull AT shako DOT sk DOT tsukuba DOT ac DOT jp>
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: GDB problem

   >My problem is that I can't get debug info attached such that GDB will
   >STEP INTO class member function which are declared/defined in various
   >header files as well as into template declarations.

   I don't think the file format produced by the current version of DJGPP
   can provide good C++ debugging information.  You'll have to try "printf
   debugging" or something similar.  Sorry.

I suppose the problem here is that class member functions defined in
the headers (ie, within the class definition) are inlined.  If you
were going to produce code that allowed this kind of function to be
"stepped into", you would need to maintain a list of all the entry and
exit points of inlined functions, since you could not simply place
breakpoints at the beginning and end of each function.

I note that one of the original poster's uses for the inlined function
was to implement a range-checking array type.  If you are going to use
such arrays a lot, that's going to be one huge list, right?

I don't know if inlining can be turned off for C++ altogether; I doubt
you want to do that anyway.  But if the set of functions that need to
be stepped into is small, you can go from the

class debuggable {

    int data;

public:

    int &access (void) { return data; }
};

style to

class debuggable {

    int data;

public:

#if !(step_into_member_functions)
    inline
#endif
    int &access (void);
};

int &debuggable::access (void) { return data; }

Ugly, yes.  But it should do what you want.

Of course, when you turn off the step_into_member_functions switch,
the function will now be inlined and this sometimes introduces bugs
(at least in earlier versions of the compiler :( ).

Another possibility that just occurred to me is that if you take the
address of the member function, the compiler must generate a
noninlined version.  This *could* be stepped into in the usual way,
but I don't know what contortions you'd have to go through to get the
debugger to actually *do* that.

Does the proposed switch to ELF help fix this situation?

-- 
Stephen Turnbull  /  Yaseppochi-gumi  /  <turnbull AT shako DOT sk DOT tsukuba DOT ac DOT jp>
http://turnbull.sk.tsukuba.ac.jp/      anon FTP: turnbull.sk.tsukuba.ac.jp
Check out Kansai-WWW, too ------------> http://pclsp2.kuicr.kyoto-u.ac.jp/