Re: [PATCH][GOLD] Use correct local symbol values in stub generation.

[=?big5?b?RG91ZyBLd2FuICjD9q62vHcp?= <dougkwan at google dot com>]

Hi,

    Attached is an updated patch.  I changed the interface to that
compute_final_local_value takes two pointers, one for input and the
other output symbol value.  I also renamed the inline function and
fixed the indentation.

-Doug

在 2010年9月8日上午8:03，Ian Lance Taylor <iant@google.com> 寫道：
> "Doug Kwan (關振德)" <dougkwan@google.com> writes:
>
>> +  if (!is_ordinary)
>> +    {
>> +      if (shndx == elfcpp::SHN_ABS || Symbol::is_common_shndx(shndx))
>> +     lv->set_output_value(lv->input_value());
>> +      else
>> +     {
>> +       this->error(_("unknown section index %u for local symbol %u"),
>> +                   shndx, r_sym);
>> +           lv->set_output_value(0);
>> +           return This::CFLV_ERROR;
>> +     }
>> +    }
>
> The indentation looks wrong here.
>
>> +  // Compute final local symbol value.  R_SYM is the index of a local
>> +  // symbol in symbol table.  LV points to a symbol value, which is
>> +  // expected to hold the input value and to be over-written by the
>> +  // final value.  SYMTAB points to a symbol table.  Some targets may want
>> +  // to know would-be-finalized local symbol values in relaxation.
>> +  // Hence we provide this method.  Since this method updates *LV, a
>> +  // callee should make a copy of the original local symbol value and
>> +  // use the copy instead of modifying an object's local symbols before
>> +  // everything is finalized.  The caller should also free up any allocated
>> +  // memory in the return value in *LV.
>> +  Compute_final_local_value_status
>> +  compute_final_local_value(unsigned int r_sym, Symbol_value<size>*lv,
>> +                         const Symbol_table* symtab);
>
> This interface seems overly complicated, because of the way it changes
> *lv.  A local symbol value is represented in two values: a number and a
> Merged_symbol_value.  Can you change this function to take a const
> Symbol_value pointer and to return those two values?
>
> Also I'm not sure the internal and external functions should have the
> same name.
>
> Also there should be a space after Symbol_value<size>*.
>
> Ian
>

Attachment: patch-cortex-a8.txt
Description: Text document