>Number:         3510
>Category:       port-i386
>Synopsis:       libm build on i386 w/i387 should be easier
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    gnats-admin (GNATS administrator)
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Fri Apr 18 04:50:05 1997
>Last-Modified:
>Originator:     Havard Eidnes
>Organization:
	SINTEF RUNIT
>Release:        NetBSD-current April 14 1997
>Environment:
	Pentium, NetBSD-current, i386, libm
System: NetBSD vader.runit.sintef.no 1.2D NetBSD 1.2D (VADER) #0: Mon Apr 14 23:27:38 MEST 1997 he@vader.runit.sintef.no:/usr/src/sys/arch/i386/compile/VADER i386


>Description:
	One should not have to edit the /usr/src/lib/libm/Makefile
	to get an architecture-optimized version of the math libraries.

	Note also that through use of a "simple SUP update", the edit
	you do in one round to fix this will be lost on the next SUP.
	(Not everyone's prepared to set up their own CVS repository.)

>How-To-Repeat:
	Observe that the i387 section is commented out in the distributed
	Makefile and has to be manually edited to enable the i387 code.

>Fix:
	The attached fix to libm/Makefile makes it possible to tell
	make via /etc/mk.conf that you have a math coprocessor on your
	system.  /etc/mk.conf might look like this:

% cat /etc/mk.conf
# System specific build options

HAS_i387=       1
% 

*** Makefile.orig	Fri Apr 18 11:35:15 1997
--- Makefile	Fri Apr 18 13:25:28 1997
***************
*** 51,54 ****
--- 51,55 ----
  .elif (${MACHINE_ARCH} == "i386")
  # Uncomment the following lines if you have a i387 (or i486 or Pentium)
+ .if defined(HAS_i387)
  .PATH:	${.CURDIR}/arch/i387
  ARCH_SRCS = e_acos.S e_asin.S e_atan2.S e_exp.S e_expf.S e_fmod.S e_log.S \
***************
*** 60,63 ****
--- 61,65 ----
  	    s_scalbn.S s_scalbnf.S s_significand.S s_significandf.S s_sin.S \
  	    s_sinf.S s_tan.S s_tanf.S
+ .endif
  .elif (${MACHINE_ARCH} == "m68k")
  .PATH:	${.CURDIR}/arch/mc68881
>Audit-Trail:
>Unformatted: