Python 1.5.2c1 -- What if long long not supported?

[Tamito Kajiyama]

Guido van Rossum <guido at CNRI.Reston.VA.US> writes:
| 
| I received a bug report that if the (unnamed) platform doesn't support
| the "long long" type, the configure script dies.  The logic in
| configure.in looks correct to me and I don't have any such Unix
| platforms handy myself...  Could someone who does investigate whether
| there is truth in this complaint?

I was able to reproduce this problem on SunOS 4.1.4_JL.

In the configure script, if the C compiler does not support the "long
long" type, the variable `ac_cv_sizeof_long_long' is not defined.  This
variable is referenced later for checking large file support, and the
test command fails due to the lack of the expected argument that follows
the second -a option.  That's why the configure script dies.

The attached is a context diff for fixing the problem.  (I'm not
familiar with autoconf, so please consider the way of fixing ;-)

diff -cr Python-1.5.2c1.orig/configure Python-1.5.2c1/configure
*** Python-1.5.2c1.orig/configure       Tue Apr 13 02:16:45 1999
--- Python-1.5.2c1/configure    Tue Apr 13 02:16:08 1999
***************
*** 2034,2039 ****
--- 2034,2040 ----
  echo $ac_n "checking for long long support""... $ac_c" 1>&6
  echo "configure:2036: checking for long long support" >&5
  have_long_long=no
+ ac_cv_sizeof_long_long=undefined
  cat > conftest.$ac_ext <<EOF
  #line 2040 "configure"
  #include "confdefs.h"
diff -cr Python-1.5.2c1.orig/configure.in Python-1.5.2c1/configure.in
*** Python-1.5.2c1.orig/configure.in    Tue Apr 13 02:10:20 1999
--- Python-1.5.2c1/configure.in Tue Apr 13 02:15:55 1999
***************
*** 352,357 ****
--- 352,358 ----
  
  AC_MSG_CHECKING(for long long support)
  have_long_long=no
+ ac_cv_sizeof_long_long=undefined
  AC_TRY_COMPILE([], [long long x; x = (long long)0;], AC_DEFINE(HAVE_LONG_LONG) have_long_long=yes)
  AC_MSG_RESULT($have_long_long)
  if test $have_long_long = yes ; then

--
KAJIYAMA, Tamito <kajiyama at grad.sccs.chukyo-u.ac.jp>