The attached patch fixes a problem where nm displays bogus information for
synthetic symbol sizes when --size-sort is used.
This happens because the synthetic symbols (dot symbols for ppc64) are
generated based on their non-dot symbols. The generation process doesn't copy
over the ELF-specific bits of the regular non-dot symbols.
When --size-sort is used, the code attempts to access the symbol size from
the ELF-specific bits and ends up reading gargabe, causing the size to be
displayed incorrectly.
With the patch, i can see dot and non-dot symbols having the same size with
--size-sort.
This doesn't fix the fact that we don't display size information for synthetic
symbols without --size-sort, which i may address in the future.
binutils/ChangeLog:
2016-12-01 Luis Machado <lgustavo@codesourcery.com>
* nm.c (sort_symbols_by_size): Don't read symbol size if symbol
is synthetic.
+2016-12-01 Luis Machado <lgustavo@codesourcery.com>
+
+ * nm.c (sort_symbols_by_size): Don't read symbol size if symbol
+ is synthetic.
+
2016-11-30 Nick Clifton <nickc@redhat.com>
PR ld/20815
asection *sec;
bfd_vma sz;
asymbol *temp;
+ int synthetic = (sym->flags & BSF_SYNTHETIC);
if (from + size < fromend)
{
sec = bfd_get_section (sym);
- if (bfd_get_flavour (abfd) == bfd_target_elf_flavour)
+ /* Synthetic symbols don't have a full type set of data available, thus
+ we can't rely on that information for the symbol size. */
+ if (!synthetic && bfd_get_flavour (abfd) == bfd_target_elf_flavour)
sz = ((elf_symbol_type *) sym)->internal_elf_sym.st_size;
- else if (bfd_is_com_section (sec))
+ else if (!synthetic && bfd_is_com_section (sec))
sz = sym->value;
else
{