Date: Sun, 9 Jun 2002 07:56:35 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: djgpp-workers AT delorie DOT com
cc: Kbwms AT aol DOT com
Subject: ISO C99 double math functions
In-Reply-To: <49.1e8b75d3.2a3236b9@aol.com>
Message-ID: <Pine.SUN.3.91.1020609074954.12685A-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

K.B. Williams kindly agreed to work on adding the missing C99 math 
functions to the DJGPP library.  However, apart of the long double 
functions, C99 also codifies several functions which return a double 
result, and which were not in the Standard C library before C99.

We have those functions in libm.a.  The question is, should we leave 
them in libm.a, move them to libc.a, or write replacement functions of 
our own and put those into libc.a?

Leaving them in libm.a means users will have to say "-lm" for some math 
functions but not for others; that can lead to confusion.  Moving them 
into libc.a means we take apart fdlibm-originated code, which will make 
it harder to upgrade to newer versions of fdlibm should we want to.  
Coming up with our own implementations means more work.

Opinions?

Here's the list of the functions we are talking about, compiled by K.B. 
Williams:

>      1: copysign
>      2: erf
>      3: erfc
>      4: fdim
>      5: fma
>      6: fmax
>      7: fmin
>      8: ilogb
>      9: lgamma
>     10: llrint
>     11: llround
>     12: logb
>     13: lrint
>     14: lround
>     15: nearbyint
>     16: nextafter
>     17: nexttoward
>     18: remainder
>     19: remquo
>     20: rint
>     21: round
>     22: scalbln
>     23: scalbn
>     24: tgamma
>     25: trunc