Re: [GOLD][PATCH] Fix ARM backend's handling of protected symbol in a DSO

[Doug Kwan (éæå) <dougkwan at google dot com>]

Both protected_1 and protected_2 fail currently on ARM.

FAIL: protected_1 (exit: 134)
=============================
protected_1: ../../../src/gold/testsuite/protected_main_1.cc:39: int main(): Ass
ertion `&f2 == get_f2_addr()' failed.

FAIL: protected_2 (exit: 134)
=============================
protected_2: ../../../src/gold/testsuite/protected_main_1.cc:39: int main(): Ass
ertion `&f2 == get_f2_addr()' failed.


On Wed, Mar 14, 2012 at 11:23 AM, Doug Kwan (關振德) <dougkwan@google.com> wrote:
> protected_1 and another protected test whose name I don't remember.  I
> can run the testsuite again but it will take a while in QEMU
>
> On Wed, Mar 14, 2012 at 11:22 AM, Ian Lance Taylor <iant@google.com> wrote:
>> "Doug Kwan (關振德)" <dougkwan@google.com> writes:
>>
>>>     I can changed that.  The code was copied from the x86_64 backend:
>>>
>>>             if (gsym->is_from_dynobj()
>>>                 || gsym->is_undefined()
>>>                 || gsym->is_preemptible()
>>>                 || (gsym->visibility() == elfcpp::STV_PROTECTED
>>>                     && parameters->options().shared())
>>>                 || (gsym->type() == elfcpp::STT_GNU_IFUNC
>>>                     && parameters->options().output_is_position_independent()))
>>>               got->add_global_with_rel(gsym, GOT_TYPE_STANDARD, rela_dyn,
>>>                                        elfcpp::R_X86_64_GLOB_DAT);
>>>
>>>
>>> The backend i386 backend also does the same.  Should we change them
>>> all to use output_is_position_independent()?
>>
>> I don't know, what's the test case?
>>
>> Ian
>>
>>> On Wed, Mar 14, 2012 at 10:00 AM, Ian Lance Taylor <iant@google.com> wrote:
>>>> "Doug Kwan (關振德)" <dougkwan@google.com> writes:
>>>>
>>>>> 2012-03-14  Doug Kwan  <dougkwan@google.com>
>>>>>
>>>>>       * gold/arm.cc (Target_arm::Scan::global): Generate R_ARM_GLOB_DAT
>>>>>       dynamic relocations for protected symbols in shared objects.
>>>>>
>>>>> Index: gold/arm.cc
>>>>> ===================================================================
>>>>> RCS file: /cvs/src/src/gold/arm.cc,v
>>>>> retrieving revision 1.146
>>>>> diff -u -u -p -r1.146 arm.cc
>>>>> --- gold/arm.cc       3 Feb 2012 20:01:01 -0000       1.146
>>>>> +++ gold/arm.cc       14 Mar 2012 08:44:14 -0000
>>>>> @@ -8314,7 +8314,9 @@ Target_arm<big_endian>::Scan::global(Sym
>>>>>           Reloc_section* rel_dyn = target->rel_dyn_section(layout);
>>>>>           if (gsym->is_from_dynobj()
>>>>>               || gsym->is_undefined()
>>>>> -             || gsym->is_preemptible())
>>>>> +             || gsym->is_preemptible()
>>>>> +             || (gsym->visibility() == elfcpp::STV_PROTECTED
>>>>> +                 && parameters->options().shared()))
>>>>>             got->add_global_with_rel(gsym, GOT_TYPE_STANDARD,
>>>>>                                      rel_dyn, elfcpp::R_ARM_GLOB_DAT);
>>>>>           else
>>>>
>>>>
>>>> Usually tests for whether to generate a dynamic reloc should check
>>>> parameters->options().output_is_position_independent() rather than
>>>> parameters->options().shared().  The difference is that the former is
>>>> also true for -pie.
>>>>
>>>> Ian