PATCH to stabsread.c:read_member_functions

PATCH to stabsread.c:read_member_functions

[Jason Merrill]

[-- Attachment #1: Type: text/plain, Size: 648 bytes --]

This patch fixes gdb.c++/classes.exp:ptype Static when used with gcc 3.x
and stabs output.  The code in question was never correct, just a likely
guess; in particular, it got things wrong even with the v2 output for
Static::ii, where the plain function name ("ii") was the same as the
mangled form of the argument list (int, int == "ii").  For v3, we don't try
to do minimal output, so it will always guess wrong.

OK?  Should I just remove the bad code rather than comment it out?

2002-01-17  Jason Merrill  <jason@redhat.com>

	* stabsread.c (read_member_functions): Never guess that the given
	physname for a static member function is a stub.


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Type: text/x-patch, Size: 691 bytes --]

*** stabsread.c.~1~	Wed Jan 16 18:52:33 2002
--- stabsread.c	Thu Jan 17 15:01:35 2002
*************** read_member_functions (struct field_info
*** 3206,3216 ****
--- 3206,3221 ----
  	    case '?':
  	      /* static member function.  */
  	      new_sublist->fn_field.voffset = VOFFSET_STATIC;
+ #if 0
+ 	      /* This was a kludge to try to get the right answer for gcc
+ 		 v2 stabs output.  In v3 we don't try to avoid writing out
+ 		 full mangled names, so this just causes problems.  */
  	      if (strncmp (new_sublist->fn_field.physname,
  			   main_fn_name, strlen (main_fn_name)))
  		{
  		  new_sublist->fn_field.is_stub = 1;
  		}
+ #endif
  	      break;
  
  	    default:

Re: PATCH to stabsread.c:read_member_functions

[Daniel Jacobowitz]

On Thu, Jan 17, 2002 at 03:06:49PM +0000, Jason Merrill wrote:
> This patch fixes gdb.c++/classes.exp:ptype Static when used with gcc 3.x
> and stabs output.  The code in question was never correct, just a likely
> guess; in particular, it got things wrong even with the v2 output for
> Static::ii, where the plain function name ("ii") was the same as the
> mangled form of the argument list (int, int == "ii").  For v3, we don't try
> to do minimal output, so it will always guess wrong.
> 
> OK?  Should I just remove the bad code rather than comment it out?
> 
> 2002-01-17  Jason Merrill  <jason@redhat.com>
> 
> 	* stabsread.c (read_member_functions): Never guess that the given
> 	physname for a static member function is a stub.
> 

Not OK, I think.  v2 -does- do minimal output, right?  We are not
dropping v2 support.

I've been meaning to fix this since I added the test.  The proper fix
might be to either:
 - know when to expect minimal output.  Does v2 ever not do it?  What
about other C++ compilers?
 - Check if it is a valid mangled name.  Will that work - can the
argument list ever be a valid mangled name in its own right?

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: PATCH to stabsread.c:read_member_functions

[Daniel Jacobowitz]

On Thu, Jan 17, 2002 at 03:06:49PM +0000, Jason Merrill wrote:
> This patch fixes gdb.c++/classes.exp:ptype Static when used with gcc 3.x
> and stabs output.  The code in question was never correct, just a likely
> guess; in particular, it got things wrong even with the v2 output for
> Static::ii, where the plain function name ("ii") was the same as the
> mangled form of the argument list (int, int == "ii").  For v3, we don't try
> to do minimal output, so it will always guess wrong.
> 
> OK?  Should I just remove the bad code rather than comment it out?
> 
> 2002-01-17  Jason Merrill  <jason@redhat.com>
> 
> 	* stabsread.c (read_member_functions): Never guess that the given
> 	physname for a static member function is a stub.
> 

Oh, and did you test the results of this on a v2/stabs testsuite?  If
not, please do so.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer