Re: [PATCH] Set non_ir_ref_dynamic if a symbol is made dynamic

["Bin.Cheng" <amker dot cheng at gmail dot com>]

On Tue, Mar 20, 2018 at 4:20 PM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Tue, Mar 20, 2018 at 9:09 AM, Christophe Lyon
> <christophe.lyon@linaro.org> wrote:
>> On 20 March 2018 at 16:39, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> On Tue, Mar 20, 2018 at 8:00 AM, Christophe Lyon
>>> <christophe.lyon@linaro.org> wrote:
>>>> On 20 March 2018 at 14:11, H.J. Lu <hjl.tools@gmail.com> wrote:
>>>>> On Tue, Mar 20, 2018 at 6:08 AM, Christophe Lyon
>>>>> <christophe.lyon@linaro.org> wrote:
>>>>>> On 20 March 2018 at 14:01, H.J. Lu <hjl.tools@gmail.com> wrote:
>>>>>>> On Tue, Mar 20, 2018 at 5:54 AM, Christophe Lyon
>>>>>>> <christophe.lyon@linaro.org> wrote:
>>>>>>>> On 20 March 2018 at 12:00, Nick Clifton <nickc@redhat.com> wrote:
>>>>>>>>> Hi H.J.
>>>>>>>>>
>>>>>>>>>> Checked into master.   Nick, is this OK to backport to 2.30 branch?
>>>>>>>>>
>>>>>>>>> Yes - please go ahead and check it in.
>>>>>>>>>
>>>>>>>>
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> Since this was checked-in in master, I've noticed
>>>>>>>> FAIL:Build pr22983
>>>>>>>> on native ARM targets.
>>>>>>>> It passes on AArch64 though.
>>>>>>>>
>>>>>>>> Can you check/fix?
>>>>>>>>
>>>>>>>
>>>>>>> I don't have arm machine.  Please show the ld log.
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> Sure, I can see:
>>>>>>
>>>>>> regexp_diff match failure
>>>>>> regexp "^ +[0-9]+: +[0-9a-f]+ +[0-9a-f]+ +FUNC +GLOBAL +DEFAULT
>>>>>> +[0-9]+ +func_attr_used_disabled$"
>>>>>> line   "     6: 00000000     0 FUNC    GLOBAL DEFAULT  UND abort@GLIBC_2.4 (2)"
>>>>>> FAIL: Build pr22983
>>>>>
>>>>> Please show "readelf --dyn-syms --wide tmpdir/pr22983"
>>>>>
>>>>
>>>> Sorry for the delay, as it was an automated build, I needed a bit of
>>>> time to reproduce it manually:
>>>> Symbol table '.dynsym' contains 9 entries:
>>>>    Num:    Value  Size Type    Bind   Vis      Ndx Name
>>>>      0: 00000000     0 NOTYPE  LOCAL  DEFAULT  UND
>>>>      1: 00000000     0 NOTYPE  WEAK   DEFAULT  UND __gmon_start__
>>>>      2: 0002102c     4 OBJECT  GLOBAL DEFAULT   23 var_attr_used_enabled
>>>>      3: 00021028     4 OBJECT  GLOBAL DEFAULT   23 var_attr_used_disabled
>>>>      4: 00000000     0 FUNC    GLOBAL DEFAULT  UND __printf_chk@GLIBC_2.4 (2)
>>>>      5: 000104f5     4 FUNC    GLOBAL DEFAULT   13 func_attr_used_enabled
>>>>      6: 00000000     0 FUNC    GLOBAL DEFAULT  UND abort@GLIBC_2.4 (2)
>>>>      7: 00000000     0 FUNC    GLOBAL DEFAULT  UND
>>>> __libc_start_main@GLIBC_2.4 (2)
>>>>      8: 000104f9     4 FUNC    GLOBAL DEFAULT   13 func_attr_used_disabled
>>>>
>>>>
>>>> So the problem is that func_attr_used_enabled and
>>>> func_attr_used_disabled do not appear next to each other.
>>>>
>>>> This does the trick:
>>>> diff --git a/ld/testsuite/ld-plugin/pr22983.d b/ld/testsuite/ld-plugin/pr22983.d
>>>> index af1d710..d75a26b 100644
>>>> --- a/ld/testsuite/ld-plugin/pr22983.d
>>>> +++ b/ld/testsuite/ld-plugin/pr22983.d
>>>> @@ -5,5 +5,6 @@ Symbol table '\.dynsym' contains [0-9]+ entries:
>>>>   +[0-9]+: +[0-9a-f]+ +[0-9a-f]+ +OBJECT +GLOBAL +DEFAULT +[0-9]+
>>>> +var_attr_used_disabled
>>>>  #...
>>>>   +[0-9]+: +[0-9a-f]+ +[0-9a-f]+ +FUNC +GLOBAL +DEFAULT +[0-9]+
>>>> +func_attr_used_enabled
>>>> +#...
>>>>   +[0-9]+: +[0-9a-f]+ +[0-9a-f]+ +FUNC +GLOBAL +DEFAULT +[0-9]+
>>>> +func_attr_used_disabled
>>>>  #...
>>>>
>>>
>>> Please try this.
>>>
>> It works for me, thanks
>>
>
> I am checking it into master branch and will backport it to 2.30 branch later.
Hi,
Should this test be skipped for bare-metal tools?  I can still see
failures like:

/.../obj/binutils/ld/../binutils/readelf --dyn-syms --wide
tmpdir/pr22983 > dump.out
extra regexps in /.../binutils-gdb/ld/testsuite/ld-plugin/pr22983.d
starting with "^Symbol table '\.dynsym' contains [0-9]+ entries:$"
FAIL: Build pr22983

"readelf --dyn-syms --wide tmpdir/pr22983" gives nothing because it
doesn't have .dynsym section?

Thanks,
bin
>
> --
> H.J.