This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Convert TEST_f_L tests from code to data
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: Andreas Jaeger <aj at suse dot com>
- Cc: <libc-alpha at sourceware dot org>
- Date: Thu, 16 May 2013 12:54:10 +0000
- Subject: Re: Convert TEST_f_L tests from code to data
- References: <Pine dot LNX dot 4 dot 64 dot 1305152017020 dot 21321 at digraph dot polyomino dot org dot uk> <51948930 dot 80405 at suse dot com>
On Thu, 16 May 2013, Andreas Jaeger wrote:
> > + /* XXX this test is incomplete. We need to have a way to specifiy
> > + the rounding method and test the critical cases. So far, only
> > + unproblematic numbers are tested. */
>
>
> Is the above comment really reflecting current state? Could you update as
> appropriate, please?
Dealing with XXX comments is already on the todo list
<http://sourceware.org/glibc/wiki/Development_Todo/Master#libm_testsuite>.
(And, we don't yet have a mechanism for tests of integer results to
specify that the result itself is ignored, as I remarked in bug 15470.)
How thorough the coverage is (beyond the rounding mode issue, which *is*
now covered) - well, that needs a much more general review for many more
functions in libm-test.inc; deciding whether coverage is sufficient isn't
necessarily simple (but the llrint tests don't cover subnormals,
infinities, NaNs, or out-of-range values at present).
--
Joseph S. Myers
joseph@codesourcery.com