🚀 go-pugleaf

In article <9209042155.AA28088@_la.tce.com> pierre@la.tce.com writes:
>I recommend testing for (m) != NULL in the ALL_MSYMBOLS macro
>in objfiles.h (gdb 4.5 at least). All other macros in objefiles.h
>are safe except this one.
>
>It happened to me that I had a bad object file and that an error
>occured in reading in dbx_symfile_init() and then gdb would core dump
>because objfile -> msymbol is NULL.

In 4.6, ALL_MSYMBOLS does test (objfile)->msymbols before calling
ALL_OBJFILE_MSYBMOLS, which actually does the iteration, so that the
remaining object files are processed:


/* Traverse all minimal symbols in one objfile.  */

#define	ALL_OBJFILE_MSYMBOLS(objfile, m) \
    for ((m) = (objfile) -> msymbols; (m)->name != NULL; (m)++)


/* Traverse all minimal symbols in all objfiles.  */

#define	ALL_MSYMBOLS(objfile, m) \
  ALL_OBJFILES (objfile)	 \
    if ((objfile)->msymbols)	 \
      ALL_OBJFILE_MSYMBOLS (objfile, m)


>So, it is safer to replace ALL_MSYMBOLS by the following :
>
>#define	ALL_MSYMBOLS(objfile, m) \
>  ALL_OBJFILES (objfile)	 \
>    for ((m) = (objfile) -> msymbols; ((m) != NULL) && ((m)->name != NULL); (m)++)

I think the 4.6 implementation is cleaner.

>Maybe an additional fix would be to remove this object file from the
>list.

Yes, but that would be nontrivial :-)

It might also be useful to notify the user when a bad object file is
found.  I'm not familiar with gdb's internals, so I can't provide a
complete fix, but here's what I have in mind:

#define	ALL_MSYMBOLS(objfile, m) \
  ALL_OBJFILES (objfile)	 \
    if (!(objfile)->msymbols)	 \
      ; /* emit warning */ \
    else \
      ALL_OBJFILE_MSYMBOLS (objfile, m)
-- 
Kevin Rodgers				kevin@traffic.den.mmc.com
Martin Marietta MS A16401		(303) 790-3971
116 Inverness Dr. East
Englewood CO 80112 USA			GO BUFFS!