In a recent note, Phil Sidler said:
> Date: Wed, 1 Mar 2000 10:08:27 -0800
>
> I found the following message on the MVS-OE mailing list in response to a
> question regarding the large size of executables in OS/390. I remade
> lynx2.8.3dev.20 setting CFLAGS = -Wl,"EDIT=NO" and it went from 5,148,672
> bytes to 2,441,216 bytes. I think this is something worth incorporating in
> the distribution.
>
I've roughly replicated Phil's results, although I was comparing
dev.20 without the CFLAGS to dev.21 with the CFLAGS.
I'm trying to figure out how to patch this. So far, I have the following:
#########################################################################
%%% Created Wed Mar 1 18:32:14 MST 2000 by target lynx.patch. %%%
diff -bru orig/lynx2-8-3/configure lynx2-8-3/configure
--- orig/lynx2-8-3/configure Fri Feb 25 05:38:35 2000
+++ lynx2-8-3/configure Wed Mar 1 18:32:05 2000
@@ -680,8 +680,12 @@
test -z "$ALL_LINGUAS" && ALL_LINGUAS=`test -d $srcdir/po && cd $srcdir/po && echo
*.po|sed -e 's/\.po//g' -e 's/*//'`
-
-
+case "$host_os" in
+os390)
+ : ${CC=c89}
+ : ${CFLAGS="-Wl,EDIT=NO"}
+ ;;
+esac
# Extract the first word of "gcc", so it can be a program name with args.
set dummy gcc; ac_word=$2
#########################################################################
But I'm pretty uncomfortable with this. I've incorporated the the other
options that I routinely supply on the command line to "configure".
But still:
o Of course, I shouldn't be patching configure, but configure.in. But I
still don't have TD's enhanced autoconf running. I suppose I could just
add similar lines to configure.in and hope for the best.
o There's a similar "case $host_os in" in configure about 5000 lines further
down. This seems misplaced to me; it happens after many of the compilation
tests, which can't possibly be run if the name of the compiler or vital
FLAGs are OS-dependent.
-- gil
--
StorageTek
INFORMATION made POWERFUL
