On Sun, 2014-06-15 at 12:50 +0200, Mark Wielaard wrote:
> On Sun, 2014-06-15 at 12:39 +0200, Kurt Roeckx wrote:
> > On Sun, Jun 15, 2014 at 12:30:02PM +0200, Mark Wielaard wrote:
> > > Another ARM oddity. A return value address in an unwind will contain an
> > > extra bit to indicate whether to return to a regular ARM or THUMB function.
> > > Add a new ebl function to return a mask to use to get the actual return
> > > address during an unwind ebl_unwind_ret_mask.
> > 
> > Does this fix all the issues on armhf now, including the
> > backtrace?
> 
> Yes, you'll need this patch, the "libebl: Add sym_func_value hook" and
> the "tests: backtrace.c accept __libc_do_syscall as first frame symname"
> patches. With libc6-dbg installed all tests should PASS with 3 SKIPs,
> without libc6-dbg installed you should see 6 SKIPs (no FAILs).
> 
> The other recent patches should be optional, but help for arches which
> don't have a fully working unwind backend and make the tests a bit more
> robust and clear when something does go wrong.

I did push the test suite cleanup patches to master now and merged the
two separate ebl_unwind_ret_mask and sym_func_value hook patches into
one using just one new ebl function for both use cases as Roland
suggested. See attached (and on the mjw/pending branch).

Cheers,

Mark