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

[Tetsuya Isaki <isaki%NetBSD.org@localhost>]

The following reply was made to PR lib/46433; it has been noted by GNATS.

From: Tetsuya Isaki <isaki%NetBSD.org@localhost>
To: gnats-bugs%NetBSD.org@localhost, jruohonen%iki.fi@localhost
Cc: lib-bug-people%netbsd.org@localhost,
        gnats-admin%netbsd.org@localhost,
        netbsd-bugs%netbsd.org@localhost
Subject: Re: lib/46433: tests/lib/libm/t_exp should not use exp() itself
Date: Wed, 09 May 2012 15:07:56 +0900

 At Wed,  9 May 2012 05:55:02 +0000 (UTC),
 Jukka Ruohonen wrote:
 > The following reply was made to PR lib/46433; it has been noted by GNATS.
 :
 >  On Wed, May 09, 2012 at 03:05:00AM +0000, isaki%NetBSD.org@localhost wrote:
 >  > for exp().  For example, this test will pass if a broken exp()
 >  > that always return zero.
 >  
 >  Do you have an opinion on how to fix this (and the related issues)?
 >     
 >  In my opinion: if a precondition for a test is not satisfied, the test
 >  should be skipped. Alternatively, we could append a "broken exp(3) check"
 >  to these tests and record expected failures on exotic things like VAX or
 >  m68k without FPU. For the first option, it would be desirable to have
 >  machine-dependent preprocessor conditionals (cf. "_HAVE_EXP_") for each and
 >  every math(3) function.
 
 How about the method to compare with the pre-calculated
 constants (like cos_angles test in t_cos.c)?
 ---
 Tetsuya Isaki <isaki%pastel-flower.jp@localhost / isaki%NetBSD.org@localhost>