Re: [PATCH] Make SIZE element for dwarf_block as size_t

[Siddhesh Poyarekar <siddhesh@redhat.com>]

On Mon, 23 Jul 2012 08:41:27 -0600, Tom wrote:
> What about dwarf2_evaluate_loc_desc_full?
> It has 'unsigned short size' as an argument but is called from
> indirect_pieced_value:
> 
>   struct dwarf2_locexpr_baton baton;
> [...]
>   return dwarf2_evaluate_loc_desc_full (TYPE_TARGET_TYPE (type),
> frame, baton.data, baton.size, baton.per_cu,
> 					piece->v.ptr.offset +
> byte_offset);

Ah yes, I missed that, thanks. I'll look through once more.

> It would have been useful to me if you had written a bit about how you
> wrote this patch and verified it.  As it is, with no information on
> that, I went grepping through the source to see if the patch was
> complete.

The fix was partly ripped out of the bitpos-expand patch[1] that I am
working on, in am attempt to make that patch a little smaller (not that
I've succeeded much at all). My current method for verification is just
to ensure that there are no regressions in the testsuite and grepping
through the code to try and ensure that I haven't missed anything
(which I did this time). I don't have a reproducer that demonstrates any
breakage due to this, but I can try to come up with one if it is
necessary.

Regards,
Siddhesh

[1] http://sourceware.org/ml/gdb-patches/2012-06/msg00851.html