Define __start/__stop symbols when there is only a dynamic def

[H.J. Lu]

On Wed, Jan 31, 2018 at 5:41 AM, Michael Matz <matz@suse.de> wrote:
> Hi,
>
> On Wed, 31 Jan 2018, H.J. Lu wrote:
>
>> On Wed, Jan 31, 2018 at 5:02 AM, Michael Matz <matz@suse.de> wrote:
>> > Hi,
>> >
>> > On Wed, 31 Jan 2018, Alan Modra wrote:
>> >
>> >> On Tue, Jan 30, 2018 at 07:56:59PM -0800, H.J. Lu wrote:
>> >> > I think we should check ref_dynamic, not def_dynamic, for __start/__stiop
>> >> > symbols.
>> >>
>> >> Yes, I've thought about it a bit more and I think that would be
>> >> better.
>> >
>> > Only checking ref_dynamic would mean that my testcase doesn't work
>> > anymore.  I.e. the pacemaker situation wouldn't be improved and we could
>> > have spared the whole merry-go-round.
>>
>> Are you really sure that your testcase:
>>
>> https://sourceware.org/ml/binutils/2018-01/msg00432.html
>>
>> doesn't have a reference from shared object?
>
> Before writing my mail I've checked that testcase, so: yes, I'm sure it
> doesn't work with just checking ref_dynamic.
>
> I've pushed the testcase and the fix approved by Alan, so you can have a
> look yourself.

Great.  I hope we finally nailed it :-).

Thanks.

-- 
H.J.