[Box Backup-dev] Solaris update

[Martin Ebourne]

On Wed, 2006-02-08 at 22:43 +0100, Stefan Norlin wrote:
> All tests pass except raidfile: 3431 failures similar to this one:
>  
> FAILURE: Condition [::memcmp(((char*)data) + seekpos[p], testbuff,
> 256) == 0] fa
> iled at testraidfile.cpp(186)
> FAILURE: Condition [readstream3.GetPosition() == pos] failed at
> testraidfile.cpp
> (185)

You know, these failures look just like the ones on the PPC platform.
Also, that only fails on the release test due to an apparent
optimisation bug.

Furthermore both platforms are big endian and the compiler bug is right
where the box_hton64 is called, which is stripped out by the
pre-processor in the big endian case.

Try the new --enable-ppc-workaround option to configure in trunk. It is
entirely likely that the optimisation bug occurs in the generic
optimisation phases of GCC before the architecture specific stuff.

Cheers,

Martin.