Don't even attempt to build the assembly-language versions of RIPEMD160

and SHA-1 when OBJFORMAT is not ELF.  Add a warning to the man page
about how SHA-1 uses bswapl, which will trap on 80386es (and the kernel
should, but doesn't currently, emulate).

lib/libmd/Makefile

@@ -1,4 +1,4 @@
-#	$Id: Makefile,v 1.29 1999/02/26 04:24:56 wollman Exp $
+#	$Id: Makefile,v 1.30 1999/02/26 18:41:45 wollman Exp $
 
 LIB=	md
 SRCS=	md2c.c md4c.c md5c.c md2hl.c md4hl.c md5hl.c \
@@ -17,6 +17,7 @@ CLEANFILES+=	md[245]hl.c md[245].ref md[245].3 mddriver \
 CFLAGS+= -I${.CURDIR}
 .PATH: ${.CURDIR}/${MACHINE_ARCH}
 
+.if ${OBJFORMAT} == "elf"
 .if exists(${MACHINE_ARCH}/sha.S)
 SRCS+=	sha.S
 CFLAGS+= -DSHA1_ASM -DELF
@@ -25,6 +26,7 @@ CFLAGS+= -DSHA1_ASM -DELF
 SRCS+=	rmd160.S
 CFLAGS+= -DRMD160_ASM -DELF
 .endif
+.endif
 
 all: md2.3 md4.3 md5.3
 

lib/libmd/sha.3

@@ -7,7 +7,7 @@
 .\" ----------------------------------------------------------------------------
 .\"
 .\" 	From: Id: mdX.3,v 1.14 1999/02/11 20:31:49 wollman Exp
-.\"	$Id: sha.3,v 1.1 1999/02/26 04:24:56 wollman Exp $
+.\"	$Id: sha.3,v 1.2 1999/02/26 18:41:45 wollman Exp $
 .\"
 .Dd February 25, 1999
 .Dt SHA 3
@@ -137,3 +137,15 @@ These functions appeared in
 No method is known to exist which finds two files having the same hash value,
 nor to find a file with a specific hash value.
 There is on the other hand no guarantee that such a method doesn't exist.
+.Pp
+The
+.Tn IA32
+(Intel) implementation of
+.Tn SHA-1
+makes heavy use of the
+.Ql bswapl
+instruction, which is not present on the original 80386.  Attempts
+to use
+.Tn SHA-1
+on those processors will cause an illegal instruction trap.
+(Arguably, the kernel should simply emulate this instruction.)