RFC: Stop using accross.m4

[Kelley Cook]

config/accross.m4 defines two macros AC_COMPILE_CHECK_SIZEOF and 
AC_C_BIGENDIAN_CROSS.

Both of these macros have more comprehensive versions that come with 
autoconf 2.59: AC_CHECK_SIZEOF and AC_C_BIGENDIAN respectively.

The first change is a nobrainer as it is a pure superset.

The second has one caveat is that in addition to WORDS_BIG_ENDIAN which 
AC_C_BIGENDIAN defines, AC_C_BIGENDIAN_CROSS defined BYTEORDER and 
HOST_WORDS_BIG_ENDIAN.  One the plus side, the macro defined 
HOST_WORDS_BIG_ENDIAN to be the same as WORDS_BIG_ENDIAN and BYTEORDER 
was either 1234 or 4321 depending on WORDS_BIG_ENDIAN.

BYTEORDER is not used within all of GCC.  HOST_WORDS_BIG_ENDIAN is only 
used in mips-tfile.c and that usage predates cvs.  As it is a target 
file, it should be using WORDS_BIG_ENDIAN.  Regardless the definition of 
the two macros are always the same .. except in one instance.

HOST_WORDS_BIG_ENDIAN was only directly modified by m32r.  I'm 95% sure 
this was due to an incorrect merge for adding m32r linux support as zack 
had previously deleted all references to this macro in 2001 and in fact 
deleted the xm-m32r file.  Nick restored this definition back into 
m32r.h and then undefines it for linux.  I deleted the 
HOST_WORDS_BIG_ENDIAN define/undef out of both of them.

In fact, looking back at it the restored version 1.8 of xm-m32r.h is 
(sans whitespace) the exact same as version 1.1. So, I'm positive that 
both xm-m32r.h and xm-linux.h should just be deleted.

http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/m32r/xm-m32r.h.diff?r1=1.1&r2=1.8&f=s

Besides would anyone actually be making a m32r->mips cross compiler 
which is the only combination that could possibly be affected?

Bootstrapped and regtested i686-pc-cygwin

OK to install
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: accross.patch
URL: <http://gcc.gnu.org/pipermail/java-patches/attachments/20050503/52151c11/attachment.ksh>