proposed fix for Autoconf "make check" problem with ksh 11/16/88g

[Paul Eggert]

Boyd Lynn Gerber lent me a test account, and I tracked the bug down to
what I think is a buffer overrun in that ancient ksh implementation.
Here is a proposed patch to work around the problem.

This patch assumes the other ksh-related patch I sent in earlier today.
(Methinks the current code hasn't been used with ksh much.  :-)

2002-03-14  Paul Eggert  <address@hidden>

        * lib/autotest/general.m4 (AT_INIT, AT_CLEANUP):
        Create at_test_all by a series of assignments,
        not by a single assignment of a long string.  The latter causes ksh
        version 11/16/88g to silently misbehave on OpenServer 5.0.6a,
        presumably because of a buffer overrun.

===================================================================
RCS file: lib/autotest/general.m4,v
retrieving revision 2.53.0.1
retrieving revision 2.53.0.2
diff -pu -r2.53.0.1 -r2.53.0.2
--- lib/autotest/general.m4     2002/03/14 13:45:38     2.53.0.1
+++ lib/autotest/general.m4     2002/03/15 01:54:51     2.53.0.2
@@ -167,7 +167,8 @@ at_groups_all='AT_groups_all'
 # numerical order.
 at_format='m4_bpatsubst(m4_defn([AT_ordinal]), [.], [.])'
 # Description of all the test groups.
-at_help_all='AT_help'])])dnl
+at_help_all=
+AT_help])])dnl
 m4_divert([OPTIONS])
 
 while test address@hidden:@] -gt 0; do
@@ -732,7 +733,8 @@ m4_define([AT_KEYWORDS],
 # Complete a group of related tests.
 m4_define([AT_CLEANUP],
 [m4_append([AT_help],
-m4_defn([AT_ordinal]);m4_defn([AT_line]);m4_defn([AT_description]);m4_ifdef([AT_keywords],
 [m4_defn([AT_keywords])])
+at_help_all=$at_help_all'm4_defn([AT_ordinal]);m4_defn([AT_line]);m4_defn([AT_description]);m4_ifdef([AT_keywords],
 [m4_defn([AT_keywords])])
+'
 )dnl
     $at_times_skip || times >$at_times_file
     )