Re: RFC: Test optimizations for multiple architectures

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

On Thu, Nov 11, 2010 at 12:24 PM, Roland McGrath <roland@redhat.com> wrote:
>> Your suggestion means:
>>
>> 1. People have to know how to set up environment.
>> 2. People have to make multiple runs of all the tests with different
>> permutations of which bits are set.
>
> No, I did not suggest that at all. ?I said you would make the tester do
> this, and by that I meant string/tester.c or a wrapper around it, or some
> functions within it, etc. ?So you, the glibc developer, have to write into
> the test code the necessary knowledge of what mask permutations are
> interesting to run.

Those internal optimized functions aren't exported from libc.so and
should remain so.  We need a way to test them. Also glibc developers
who write string/memory testcases may not know any mask permutations
at all.

My approach leaves which optimized functions to test to developers
who work on those functions via the __libc_func interface. If you are
writing string/memory testcases, you don't have to know any mask
permutations.  All optimized functions which are exported via
the __libc_func interface will be tested automatically.

Please checkout hjl/test-ifunc branch at

http://git.kernel.org/?p=devel/glibc/hjl/x86.git;a=summary

to try it out.  Any feedbacks are appreciated.

Thanks.


-- 
H.J.