patches: config for m68k-apollo-bsd*, or not

patches: config for m68k-apollo-bsd*, or not

[Nathanael Nerode]

[-- Attachment #1: Type: text/plain, Size: 802 bytes --]

All right.  It appears that out of all of gcc and src, m68k-apollo-bsd* 
(as opposed to m68k-apollo-sysv*) is supported potentially only by gas.  

I am strongly inclined to kill the top level configuration logic for it, 
because it is causing trouble in autoconfiscation.  This is the 
deapollo.diff patch.  I would like it approved for the gcc 3.4 basic 
improvements branch and subsequently for src when that branch lands.

If there are strong objections, the alternate patch apollo.diff updates 
the top level logic to allow the top level Makefile.in to be made 
autoconf-compatible.  But I prefer the first patch as it's simpler.

	* configure.in (m68k-apollo-bsd*): Drop support.

-or-
	* configure: Add hook for per-build-machine override of BISON.
	* configure.in: Use it for m68k-apollo-bsd*.

[-- Attachment #2: deapollo.diff --]
[-- Type: text/plain, Size: 1110 bytes --]

Index: configure.in
===================================================================
RCS file: /cvs/gcc/gcc/configure.in,v
retrieving revision 1.177.4.5
diff -u -r1.177.4.5 configure.in
--- configure.in	10 Nov 2002 21:14:18 -0000	1.177.4.5
+++ configure.in	10 Nov 2002 21:26:46 -0000
@@ -943,16 +943,6 @@
     tentative_cc="cc -A ansi -A runtype,any -A systype,any -U__STDC__ -DUSG"
     host_makefile_frag="config/mh-apollo68"
     ;;
-  m68k-apollo-bsd*)
-    #None of the Apollo compilers can compile gas or binutils.  The preprocessor
-    # chokes on bfd, the compiler won't let you assign integers to enums, and
-    # other problems.  Defining CC to gcc is a questionable way to say "don't use
-    # the apollo compiler" (the preferred version of GCC could be called cc,
-    # or whatever), but I'm not sure leaving CC as cc is any better...
-    #CC=cc -A ansi -A runtype,any -A systype,any -U__STDC__ -DNO_STDARG
-    tentative_cc=gcc
-    host_makefile_frag="config/mh-a68bsd"
-    ;;
   m88k-dg-dgux*)
     tentative_cc="gcc -Wall -ansi -D__using_DGUX"
     host_makefile_frag="config/mh-dgux"

[-- Attachment #3: apollo.diff --]
[-- Type: text/plain, Size: 1301 bytes --]

? aix.diff
? apollo.diff
? p1.patch
? p2.patch
? p3.patch
? p4.patch
Index: configure
===================================================================
RCS file: /cvs/gcc/gcc/configure,v
retrieving revision 1.46.6.3
diff -u -r1.46.6.3 configure
--- configure	10 Nov 2002 21:14:18 -0000	1.46.6.3
+++ configure	10 Nov 2002 21:22:26 -0000
@@ -1037,7 +1037,8 @@
   AR_FOR_TARGET=${AR_FOR_TARGET}${extra_arflags_for_target}
   AS=as
   AS_FOR_TARGET='\$(USUAL_AS_FOR_TARGET)'
-  BISON='$(USUAL_BISON)'
+  # default to previously set bison: workaround for m68k-apollo-bsd
+  BISON=${BISON-'\$(USUAL_BISON)'}
   CC_FOR_BUILD='\$(CC)'
   DLLTOOL=dlltool
   DLLTOOL_FOR_TARGET='\$(USUAL_DLLTOOL_FOR_TARGET)'
Index: configure.in
===================================================================
RCS file: /cvs/gcc/gcc/configure.in,v
retrieving revision 1.177.4.5
diff -u -r1.177.4.5 configure.in
--- configure.in	10 Nov 2002 21:14:18 -0000	1.177.4.5
+++ configure.in	10 Nov 2002 21:22:32 -0000
@@ -103,6 +103,12 @@
 
 # per-target:
 
+# This is a irritating workaround for m68k-apollo-bsd,
+# which wants to use yacc for bison.
+case ${build} in
+  m68k-apollo-bsd*) BISON=yacc ;;
+esac
+
 # Define is_cross_compiler to save on calls to 'test'.
 is_cross_compiler=
 if test x"${host}" = x"${target}" ; then

Re: patches: config for m68k-apollo-bsd*, or not

[Zack Weinberg]

On Sun, Nov 10, 2002 at 04:33:00PM -0500, Nathanael Nerode wrote:
[...]

> +# This is a irritating workaround for m68k-apollo-bsd,
> +# which wants to use yacc for bison.
> +case ${build} in
> +  m68k-apollo-bsd*) BISON=yacc ;;
> +esac

Given that we have bison-specific code in most of our parsers at this
point, I think it's correct to nuke everything that sets BISON to
something else.

zw

Re: patches: config for m68k-apollo-bsd*, or not

[Nathanael Nerode]

DJ Delorie wrote:
> FYI Removing targets isn't up to me...
> 
Yeah... I don't really know who should approve it, given that it's one 
of those things which isn't really supported in any of the 
subdirectories but is supported at the top level. :-P  Global maintainer 
please?

Zack Weinberg thought the "support" for m68k-apollo-bsd* was clearly 
nonfunctional, so I think it's all right to make this change; I just 
need approval from someone who can give approval.  There's no actual 
specific support for the 'bsd' variant in any subdirectories, while 
there is support for the 'sysv' variant in several places and the 
unspecified m68k-apollo* version in some places.

--Nathanael