Re: FAIL: Build pr22983 (both x86-64 and aarch64, Gentoo)

[Nick Clifton <nickc at redhat dot com>]

Hi Alan,

>> 	PR 22983
>> 	* testsuite/ld-plugin/lto.exp: Use individual tests to check for
>> 	the presence of each expected symbol.

> We are likely going to see more of this sort of problem.  The order of
> symbols in .dynsym depends on a hash table traversal over the main
> linker symbol table.  (See _bfd_elf_link_renumber_dynsyms.)  Since the
> table entry occupied by a symbol depends on its hash and the table
> size, the order can change with different symbol versions (not the
> case here as we aren't testing libc symbols) or when the table is
> resized.  I suspect the latter has occurred due to changes in the
> system libraries.

I did wonder if it would be worth adding a "sort-symbols" option to readelf
so that we could guarantee their order in the output.  Either that or we
could change the linker so that it provides some extra guarantees to the
order of the symbols in the dynamic symbol table.  I did not particularly
like that second option though.

Thoughts ?

Cheers
  Nick