[PATCH 2/4] nm: fall back to heuristic when ELF symbol has zero size

[Jan Beulich]

Size being set for a symbol isn't a strict requirement in ELF. For ones
not having their size set, fall back to the same logic as used for non-
ELF, non-COFF symbols.

While there switch to using elf_symbol_from() instead of kind of open-
coding it.
---
In the testsuite the non-format-specific test (not setting symbol size,
nm-1.s) may want to be run in all cases. Doing so would, however, mean
either quite a bit of code duplication in nm.exp, or quite a bit of
churn there. An I'm not sure my expect is good enough to arrange for the
latter, i.e. putting the respective testing in a loop.

--- a/binutils/nm.c
+++ b/binutils/nm.c
@@ -80,6 +80,7 @@ struct extended_symbol_info
 #define SYM_STAB_OTHER(sym)  (sym->sinfo->stab_other)
 #define SYM_SIZE(sym) \
   (sym->elfinfo \
+   && sym->elfinfo->internal_elf_sym.st_size \
    ? sym->elfinfo->internal_elf_sym.st_size \
    : sym->coffinfo \
      && ISFCN (sym->coffinfo->native->u.syment.n_type) \
@@ -1088,6 +1089,7 @@ sort_symbols_by_size (bfd *abfd, bool is
       asection *sec;
       bfd_vma sz;
       asymbol *temp;
+      const elf_symbol_type *elfsym;
       const coff_symbol_type *coffsym;
 
       if (from + size < fromend)
@@ -1108,8 +1110,9 @@ sort_symbols_by_size (bfd *abfd, bool is
 	 we can't rely on that information for the symbol size.  Ditto for
 	 bfd/section.c:global_syms like *ABS*.  */
       if ((sym->flags & (BSF_SECTION_SYM | BSF_SYNTHETIC)) == 0
-	  && bfd_get_flavour (abfd) == bfd_target_elf_flavour)
-	sz = ((elf_symbol_type *) sym)->internal_elf_sym.st_size;
+	  && (elfsym = elf_symbol_from (sym)) != NULL
+	  && elfsym->internal_elf_sym.st_size != 0)
+	sz = elfsym->internal_elf_sym.st_size;
       else if ((sym->flags & (BSF_SECTION_SYM | BSF_SYNTHETIC)) == 0
 	       && (coffsym = coff_symbol_from (sym)) != NULL
 	       && ISFCN (coffsym->native->u.syment.n_type)