Re: lib/46433: tests/lib/libm/t_exp should not use exp() itself

[christos%zoulas.com@localhost (Christos Zoulas)]

On May 30,  4:00pm, jruohonen%iki.fi@localhost (Jukka Ruohonen) wrote:
-- Subject: Re: lib/46433: tests/lib/libm/t_exp should not use exp() itself

| The following reply was made to PR lib/46433; it has been noted by GNATS.
| 
| From: Jukka Ruohonen <jruohonen%iki.fi@localhost>
| To: Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>
| Cc: gnats-bugs%NetBSD.org@localhost
| Subject: Re: lib/46433: tests/lib/libm/t_exp should not use exp() itself
| Date: Wed, 30 May 2012 18:57:08 +0300
| 
|  On Thu, May 31, 2012 at 12:47:10AM +0900, Izumi Tsutsui wrote:
|  > > And once again, the idea has never been to skip tests but to report these
|  > > as expected failures. As a developer, you should benefit from this as you
|  > > see directly whether and when a bug is fixed. Thus, this particular case
|  > > still requires the MD-specific check, which points to this PR upon 
failure.
|  > 
|  > Only if the test correctly fails on bogus implementation?
|  >
|  > Tests which could return false positive won't show if a bug is fixed
|  > and just hide bugs.
|  
|  Obviously. No one has claimed that there wouldn't be bugs in the tests. For
|  this case, there is still the twin PR lib/46434.
|  
|  But when trying to see the forest from the trees: the decision is not always
|  clear particularly when emulators come to the picture. For instance, we've
|  captured a fair amount of Qemu-related FP bugs. But now that these are
|  counted as expected failures, we've noticed that new Qemu releases have
|  fixed some of these issues; see PR misc/44767.
|  
|  And I don't particularly like the current #ifdef __vax__ madness either...

That should be changed to HAVE_IEEE_MATH or something. I think I started
doing that a while ago.

christos