https://sourceware.org/bugzilla/show_bug.cgi?id=27659
--- Comment #16 from Alan Modra <amodra at gmail dot com> ---
(In reply to Nick Clifton from comment #12)
> It may be that a NULL shndx pointer is
> actually allowed
No, it isn't allowed (and passing a 0 as the *pointer* arg to H_PUT_32, see
bfd_putl32/bfd_putb32, will surely result in a segfault).
The patch was commited with 6e0dfbf420, accidentally I assume..
If you have an internal st_shndx greater than a little less than 64k, then ELF
requires those symbols to be represented externally with st_shndx 0xffff and
have the actual section index stored in an array of 32-bit words in a
SHT_SYMTAB_SHNDX section. See elflink.c:bfd_elf_final_link
/* The real buffer will be allocated in elf_link_swap_symbols_out. */
flinfo.symshndxbuf
= (elf_numsections (abfd) > (SHN_LORESERVE & 0xFFFF)
? (Elf_External_Sym_Shndx *) -1 : NULL);
I think the most likely cause of this problem is a symbol st_shndx entry being
overwritten with bogus data somehow.
--
You are receiving this mail because:
You are on the CC list for the bug.
