Re: [PATCH 1/5] configure: check for libipt

[Pedro Alves <palves@redhat.com>]

On 06/30/2015 03:54 PM, Metzger, Markus T wrote:

>> Is the library host independent?  That is, does it handle
>> host vs target endianness, integer types, etc.?  E.g., is a big endian PPC
>> host debugging against an x86-64 gdbserver able to use libipt?  Another
>> example would be a big endian PPC host loading an x86-64 core dump that
>> includes ipt data (once we get to it).
> 
> No.  It may not be too hard to make it work, though.  Most of the memory
> accesses are byte-wise, already.
> 
> I don't have a PPC system to test this configuration.  

FYI, there's a few quite powerful ppc64 machines in the
gcc compile farm (gcc110-gcc112):

  https://gcc.gnu.org/wiki/CompileFarm

> Is this something
> where Sergio's buildbot can help?  How is this cross-platform testing
> usually handled?

Probably not, as no buildslave configured in our buildbot does any cross testing.
It wouldn't be impossible, we could use the gcc compile farm machines for
that, just nobody ever set it up.

> Until libipt is available for PPC, you simply can't build a GDB with Intel PT
> support on PPC.  When you configure GDB, HAVE_LIBIPT will be undefined
> and GDB will fall back to BTS or report an error.  We're not breaking anything.
> The feature will just not be available on all platforms.

OK, that's fine.  Was just checking.

Thanks,
Pedro Alves