Re: [Libunwind-devel] [PATCH 7/7] Implement _UPTi_find_unwind_table() for ARM

[Anderson Lizardo]

ext ext Daniel Jacobowitz wrote:
> On Wed, Jun 25, 2008 at 08:58:30AM -0400, Anderson Lizardo wrote:
>> We have no way to do tests on other architectures other than x86 and ARM
>> I'm afraid, and some other architectures already seem to have their own
>> _UPTi_find_unwind_table() implementation, notably:
>>
>> x86
>> x86_64
>> HPPA
>> ppc64
>> ia64
>>
>> So we added this new _UPTi_find_unwind_table() as an ARM only function
>> for now, but it can be easily used by other architectures if anyone
>> implement the other necessary bits for them (I think at least the
>> register offsets are necessary).
>>
>> Sounds reasonable?
> 
> IIUC, what you have is a routine to find unwind tables by looking in
> .debug_frame while remote debugging.  If that was a function not
> conditioned on ARM, other platforms would be free to share it.  That
> may be better.

But what about those architectures that can find the unwind tables by other
ways (such as x86, which currently uses eh_frame)? Should we do something
like looking for eh_frame first then, if not found, try debug_frame? Sounds
the best solution IMHO.

What about IA64? sounds like it is handled differently in a lot of places,
so I'd keep it's separate implementation for now, and use the new "unified"
_UPTi_find_unwind_table() version for the other architectures.

Regards,
-- 
Anderson Lizardo
Instituto Nokia de Tecnologia (INdT)
Manaus - Brazil