Re: PATCH: Add x32 abilist files

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Tue, May 15, 2012 at 3:32 PM, Joseph S. Myers
<joseph@codesourcery.com> wrote:
> On Tue, 15 May 2012, Roland McGrath wrote:
>
>> 1. Put the x86_64.* patterns back into the main file, and use
>> ? ?%ifdef __ILP32__
>> ? ?to distinguish x32. ?But if the i686 compiler defines __ILP32__ too,
>> ? ?then this will get the wrong LD_SO. ?However, that needs a solution
>> ? ?anyway, since for i686/x86_64/x32 the one <gnu/lib-names.h> is supposed
>> ? ?to be triarch.
>
> This requires a macro defined in configure, rather than one predefined by
> the compiler, as with the ARM hard-float dynamic linker issue.
>
>> 2. Put them into a new common place like sysdeps/x86/shlib-versions.
>> ? ?Then make sysdeps/{x86_64,i386}/Implies list x86.
>
> I like the sysdeps/x86 approach.
>
>> Unless we figure that out in the next few hours, I think the right thing
>> for now is to revert HJ's commit so that the i686 build does not regress
>> before we resolve this properly.
>
> I don't think this is particularly significant as a regression - I'm
> pretty sure there are various (non-generated) headers installed for i686
> that aren't biarch-capable at all, so for biarch-capable headers you
> actually need to take those installed for x86_64 at present.

sysdeps/x86 won't solve x32 problem since scripts/lib-names.awk
doesn't support more than 2  ABIs.

> But I like sysdeps/x86 (and sysdeps/unix/sysv/linux/x86, etc.) as a
> solution for this issue - it would be best for all three ABIs to install
> exactly the same set of headers, which would work for all three ABIs,
> rather than having multiple copies of various installed headers, some
> versions of a header being biarch-capable and some not. ?(There are some
> files other than installed headers that might be shared between x86 and
> x86_64 as well.)
>

That is a good idea.

-- 
H.J.