> Yes, I executed the test suite on the following systems:
> Ubuntu 12.04 64bit
> Ubuntu 14.04 64bit
> Fedora 17 64bit
> 
> No regressions on these systems. However, I see some flaky tests with and without my patch.
> 
> Any other recommended systems I should try?

No, that's good  enough. Some tests are indeed still a little flaky
for me too (Eg: attach-many-short-lived-threads.exp).

Here is an updated version to your patch. Here are the list of
changes I made:

  . I fixed the formatting in a number of function calls where
    the subsequent parameters where misaligned. Eg:

       |   temp = obstack_alloc (&objfile->objfile_obstack,
       |-    sizeof (struct dynamic_prop_list));
       |+                        sizeof (struct dynamic_prop_list));

    (the second argument should be aligned to the first one).

  . I simplified add_dyn_prop by using obstack_copy.

  . I rewrote copy_dynamic_prop_list in a simpler way. But I'd like
    you to review my implementation, in case there is a flaw in my
    logic.

  . I swapped the order of the parameters in copy_dynamic_prop_list.

  . I added more documentation to to the various types we are adding.
    It's fairly trivial, but allows us to more closely follow the
    project's guidelines.

If you are happy with that, and you agree with copy_dynamic_prop_list's
new implementation, then why don't you write the revision log and
ChangeLog and resubmit. I'll take one last look and we can move on
to the next phase.

I've tested this version on x86_64-linux, no regression.

-- 
Joel