[Bug ld/10629] GNU ld (GNU Binutils) 2.20.51.20090910 cores

[george dot tovrea at baesystems dot com]

------- Additional Comments From george dot tovrea at baesystems dot com  
2009-09-15 15:32 -------
Subject: RE:  GNU ld (GNU Binutils) 2.20.51.20090910 cores

More info: The following is a snippet of the code where the patch was applied 
and I added debug statements. I added the output (see >>) where the failure 
appears to be occuring

  if (!info->shared)
    {
      const char *name;

// DEBUG DEBUG
          fprintf(stderr, "DEBUG DEBUG: I AM AT  %s, line %d \n", __FILE__, 
__LINE__ );
// DEBUG DEBUG

      name = SGI_COMPAT (abfd) ? "_DYNAMIC_LINK" : "_DYNAMIC_LINKING";
      bh = NULL;
      if (!(_bfd_generic_link_add_one_symbol
            (info, abfd, name, BSF_GLOBAL, bfd_abs_section_ptr, 0,
             NULL, FALSE, get_elf_backend_data (abfd)->collect, &bh)))
        return FALSE;

// DEBUG DEBUG
          fprintf(stderr, "DEBUG DEBUG: I AM AT  %s, line %d \n", __FILE__, 
__LINE__ );
// DEBUG DEBUG

      h = (struct elf_link_hash_entry *) bh;
      h->non_elf = 0;
      h->def_regular = 1;
      h->type = STT_SECTION;

// DEBUG DEBUG
          fprintf(stderr, "DEBUG DEBUG: I AM AT  %s, line %d \n", __FILE__, 
__LINE__ );
// DEBUG DEBUG

      if (! bfd_elf_link_record_dynamic_symbol (info, h))
        return FALSE;

// DEBUG DEBUG
          fprintf(stderr, "DEBUG DEBUG: I AM AT  %s, line %d \n", __FILE__, 
__LINE__ );
// DEBUG DEBUG

>> DEBUG DEBUG: I AM AT  elfxx-mips.c, line 7003

      if (! mips_elf_hash_table (info)->use_rld_obj_head)
        {
// DEBUG DEBUG
          fprintf(stderr, "DEBUG DEBUG: I AM AT  %s, line %d \n", __FILE__, 
__LINE__ );
// DEBUG DEBUG

>> THE PRECEEDING LINE WAS NOT PRINTED

          /* __rld_map is a four byte word located in the .data section
             and is filled in by the rtld to contain a pointer to
             the _r_debug structure. Its symbol value will be set in
             _bfd_mips_elf_finish_dynamic_symbol.  */
          s = bfd_get_section_by_name (abfd, ".rld_map");

// DEBUG DEBUG
          fprintf(stderr, "DEBUG DEBUG: I AM AT  %s, line %d \n", __FILE__, 
__LINE__ );
// DEBUG DEBUG

          if (s != NULL)
            {
                name = SGI_COMPAT (abfd) ? "__rld_map" : "__RLD_MAP";
                bh = NULL;
                if (!(_bfd_generic_link_add_one_symbol
                      (info, abfd, name, BSF_GLOBAL, s, 0, NULL, FALSE,
                       get_elf_backend_data (abfd)->collect, &bh)))
                    return FALSE;

                h = (struct elf_link_hash_entry *) bh;
                h->non_elf = 0;
                h->def_regular = 1;
                h->type = STT_OBJECT;

                if (! bfd_elf_link_record_dynamic_symbol (info, h))
                    return FALSE;
            }
        }
    }


-----Original Message-----
From: Tovrea, George W (US SSA) 
Sent: Tuesday, September 15, 2009 8:12 AM
To: 'address@hidden'
Subject: RE: [Bug ld/10629] GNU ld (GNU Binutils) 2.20.51.20090910 cores

I tried the patch without success but the problem does appear to be in that 
area of the code. I inserted some debug statements and found only one was hit 
before the core. I am trying to narrow it down and I'll keep you up to date on 
what I find.

Regards,
Bill T.

-----Original Message-----
From: nickc at redhat dot com [mailto:address@hidden 
Sent: Monday, September 14, 2009 9:26 AM
To: Tovrea, George W (US SSA)
Subject: [Bug ld/10629] GNU ld (GNU Binutils) 2.20.51.20090910 cores


------- Additional Comments From nickc at redhat dot com  2009-09-14 16:26 
-------
Hi Bill,

> Not sure why you need simple.o since it is "simple" to compile,

Because without it I have to create a compiler and then compile simple.cc.  Plus
I have to hope that I am using the same set of compiler sources as you and that
I have configured the compiler in the same way.

> but it and xerces 28 are attached

Well the simple.o file appears to be missing, but I had a go at creating one for
myself.  As a result I have created a workaround patch which *may* work.  Please
give it a go and see if it helps.  I am not familiar with this particular area
of the linker though, so this is definitely a workaround not a fix.

Cheers
  Nick




-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=10629

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.