Re: [PATCH] BFD vector for elf32-i386-nacl

[Roland McGrath <mcgrathr at google dot com>]

elf_i386_link_hash_table is slightly overloaded for such uses (the
is_vxworks and plt0_pad_byte fields), but IMHO that is a bad kludge.  The
main purpose of that structure is for actual per-object data specific to
the backend.  What I've added is constant per-backend data that will never
vary across objects and it would be wasteful to duplicate it in the struct
elf_i386_link_hash_table for each individual object.


Thanks,
Roland


On 7/26/11, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Tue, Jul 26, 2011 at 1:56 PM, Roland McGrath <mcgrathr@google.com> wrote:
>> This patch adds the BFD target vector for the new target elf32-i386-nacl.
>> This is to support Native Client (http://code.google.com/p/nativeclient/),
>> which is a new pseudo-operating system that uses vanilla ELF but has a
>> significantly nonstandard ABI. ?(It is already known to config.sub as
>> 'nacl'.)
>>
>> This is far from being complete support for this target, but it should
>> be everything that will be needed in BFD proper. ?It's gotten only the
>> lightest testing, and of course the target-specific details are subject
>> to potential fixes or changes later.
>>
>> I'm floating this early primarily to get feedback about the
>> parameterization I've added to the shared elf32-i386.c code.
>> The approach I took seems pretty clean to me, and moreso than
>> e.g. the way vxworks was handled.
>>
>> I have tested that the i686-linux target still passes check-binutils and
>> check-ld. ?I'm not sure that tests much about correct PLT generation,
>> which is the only area of the shared code really touched here. ?If there
>> are other good ways to test that (short of rebuilding a whole OS with
>> this linker), please let me know.
>>
>> Ok for trunk?
>>
>> Thanks,
>> Roland
>>
>>
>> bfd/
>> 2011-07-25 ?Roland McGrath ?<mcgrathr@google.com>
>>
>> ? ? ? ?* elf32-i386.c (NACL_PLT_ENTRY_SIZE, NACLMASK): New macros.
>> ? ? ? ?(elf_i386_nacl_plt0_entry): New variable.
>> ? ? ? ?(elf_i386_plt_entry): New variable.
>> ? ? ? ?(elf_i386_nacl_pic_plt0_entry): New variable.
>> ? ? ? ?(elf_i386_nacl_pic_plt_entry): New variable.
>> ? ? ? ?(elf_i386_nacl_arch_bed): New variable.
>> ? ? ? ?(elf_backend_arch_data): New macro setting for elf_i386_nacl_vec
>> stanza.
>> ? ? ? ?(elf_backend_plt_alignment): Likewise.
>>
>> ? ? ? ?* elf32-i386.c (struct elf_i386_backend_data): New type.
>
> We have elf_i386_link_hash_table for similar purpose.  Is that
> possible to put elf_i386_backend_data fields there?
>
>
> H.J.
>