Re: [RFA] Fix a memory corruption in mdebugread.c

[Joel Brobecker <brobecker@gnat.com>]

[-- Attachment #1: Type: text/plain, Size: 542 bytes --]

> OK.  Assuming that the name of the class's stEnd is not mangled like a
> constructor, my worry is unfounded.  If you add a comment to that effect
> then this patch is OK.

Class names appear not to be mangled, so indeed there should be no
name "clash". I checked the following change in (I expanded the comment
to explain that a method can not have the same name as the class name):

2003-01-03  J. Brobecker  <brobecker@gnat.com>

        * mdebugread.c (parse_symbol): Count until the stEnd matching
        the structure name.

-- 
Joel

[-- Attachment #2: mdebugread.c.diff --]
[-- Type: text/plain, Size: 1891 bytes --]

Index: mdebugread.c
===================================================================
RCS file: /cvs/src/src/gdb/mdebugread.c,v
retrieving revision 1.32
diff -c -3 -p -r1.32 mdebugread.c
*** mdebugread.c	17 Dec 2002 00:39:07 -0000	1.32
--- mdebugread.c	3 Jan 2003 13:59:00 -0000
*************** parse_symbol (SYMR *sh, union aux_ext *a
*** 865,871 ****
  	    switch (tsym.st)
  	      {
  	      case stEnd:
! 		goto end_of_fields;
  
  	      case stMember:
  		if (nfields == 0 && type_code == TYPE_CODE_UNDEF)
--- 865,888 ----
  	    switch (tsym.st)
  	      {
  	      case stEnd:
!                 /* C++ encodes class types as structures where there the
!                    methods are encoded as stProc. The scope of stProc
!                    symbols also ends with stEnd, thus creating a risk of
!                    taking the wrong stEnd symbol record as the end of
!                    the current struct, which would cause GDB to undercount
!                    the real number of fields in this struct.  To make sure
!                    we really reached the right stEnd symbol record, we
!                    check the associated name, and match it against the
!                    struct name.  Since method names are mangled while
!                    the class name is not, there is no risk of having a
!                    method whose name is identical to the class name
!                    (in particular constructor method names are different
!                    from the class name).  There is therefore no risk that
!                    this check stops the count on the StEnd of a method.  */
!                 if (strcmp (debug_info->ss + cur_fdr->issBase + tsym.iss,
!                             name) == 0)
!                   goto end_of_fields;
!                 break;
  
  	      case stMember:
  		if (nfields == 0 && type_code == TYPE_CODE_UNDEF)