This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH][GOLD] Use offsets within output sections during a relocatable link.
I used uint64_t because START and Output_section::address() are of
type uint64_t. I can certainly change that to Address.
-Doug
在 2010年5月27日上午1:32,Ian Lance Taylor <iant@google.com> 寫道:
> "Doug Kwan (關振德)" <dougkwan@google.com> writes:
>
>> 2010-05-27 Doug Kwan <dougkwan@google.com>
>>
>> * object.cc (Sized_relobj::do_finalize_local_symbols): Use offset
>> from start of output section instead of address for a local symbol
>> in a merged or relaxed section when doing a relocatable link.
>
> Thanks for finding this.
>
>
>> --- gold/object.cc 26 May 2010 15:47:39 -0000 1.123
>> +++ gold/object.cc 26 May 2010 15:57:55 -0000
>> @@ -1827,7 +1827,12 @@ Sized_relobj<size, big_endian>::do_final
>> const Output_section_data* posd =
>> os->find_relaxed_input_section(this, shndx);
>> if (posd != NULL)
>> - lv.set_output_value(posd->address());
>> + {
>> + uint64_t relocatable_link_adjustment =
>> + relocatable ? os->address() : 0;
>> + lv.set_output_value(posd->address()
>> + - relocatable_link_adjustment);
>> + }
>
> This is a template function. The variable relocation_link_adjustment
> should have type Address (a typedef in Sized_relobj), not uint64_t.
>
>
>> @@ -1835,9 +1840,14 @@ Sized_relobj<size, big_endian>::do_final
>> {
>> // We have to consider the addend to determine the
>> // value to use in a relocation. START is the start
>> - // of this input section.
>> + // of this input section. If we are doing a relocatable
>> + // link, use offset from start output section instead of
>> + // address.
>> + uint64_t adjusted_start =
>> + relocatable ? start - os->address() : start;
>> Merged_symbol_value<size>* msv =
>> - new Merged_symbol_value<size>(lv.input_value(), start);
>> + new Merged_symbol_value<size>(lv.input_value(),
>> + adjusted_start);
>> lv.set_merged_symbol_value(msv);
>> }
>> }
>
> Here too, for the variable adjusted_start.
>
> This is OK with those changes.
>
> Thanks.
>
> Ian
>