Re: PATCH: Add x32 support to dynamic linker audit

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

On Fri, Mar 23, 2012 at 8:04 AM, Carlos O'Donell
<carlos@systemhalted.org> wrote:
> On Thu, Mar 22, 2012 at 8:08 PM, Roland McGrath <roland@hack.frob.com> wrote:
>>> (a) Use `struct La_x32_*;' in ldsodefs.h even though it doesn't
>>> actually match the real type used, but this doesn't matter because
>>> when compiling for the right machine it is changed by the macros and
>>> works out. It's only when *not* compiling for x32 that the debug info
>>> won't match, but you also don't get the proper structure definition
>>> since bits/link.h comes from the other machine you're targetting.
>>>
>>> (b) Use `#define La_x32_*' in ldsodefs.h, mirroring what's in
>>> bits/link.h for x86_64. The downside here is that this is a
>>> duplication of code.
>>>
>>> (c) Rewrite ldsodefs.h to stop listing all of the interface for all of
>>> the machines.
>>>
>>> (d) Rewrite bits/link.h to expose a real structure called La_x32_*
>>> when compiling for x32, otherwise La_x86_64_* when compiling for
>>> x86_64.
>>>
>>> I think that (a) is clean, works correctly when compiling for the
>>> right machine, and is the simplest solution.
>>>
>>> Having said that I'd love to see (c), but I feel bad heaping that work on H.J.
>>
>> I think (a) is adequately clean for now.
>> I think (d) is arguably cleaner than the status quo just on its own terms.
>> I think (c) makes plenty of sense but we haven't heard any perspectives
>> on why it was originally done the way it was.
>> I think (b) is an unworthy kludge.
>
> Thanks for the feedback.
>
> H.J. could you please checkin your fix for (a)?

Done.

Thanks.

-- 
H.J.