Not relevant to 21.4.
Sun hides utilities such as ar; this patch makes configure barf,
rather than letting you get all the way to installation.
The second hunk in configure.ac is a literal translation of the
Autoconf 2.13 macro to 2.5x. I wonder if maybe `long' wouldn't be a
better choice? Is it worth doing up config.h.in? Nobody has ever
diff --git a/ChangeLog b/ChangeLog
index 6ee200d..5de4caf 100644
＠＠ -0,0 +1,5 ＠＠
+2006-12-24 Stephen J. Turnbull <stephen(a)xemacs.org>
+ * configure.ac (ar): Sun has it but hides it; check for it.
+ (ssize_t): Use the modern check and document todo.
diff --git a/configure.ac b/configure.ac
index 15facd5..abeb1fa 100644
＠＠ -2877,6 +2877,13 ＠＠ dnl Some other nice autoconf tests. If you add a test here which
dnl should make an entry in src/config.h, do not forget to add an
dnl #undef clause to src/config.h.in for autoconf to modify.
+dnl Sun likes to hide these utilities. Steve Baur asked that we make
+dnl the failure occur at configure time, not at use time.
+if test "$AR" = "not_found"; then
+ XE_DIE(["ar not found -- is PATH set correctly?"])
+dnl These tests don't necessarily check that the program exists.
＠＠ -2937,8 +2944,13 ＠＠ AC_TYPE_PID_T
-dnl #### deprecated 2.13-ism
+dnl #### Translated from deprecated 2.13-ism.
+dnl We probably want to DTRT with #ifdef HAVE_SSIZE_T in src/config.h.in.
+dnl Specifically, we should typedef st sizeof(ssize_t) == sizeof(size_t).
+ [AC_DEFINE_UNQUOTED([ssize_t], [int],
+ [Define to `int' if
+ <sys/types.h> does not define.])])
dnl not AC_CHECK_TYPE; lisp.h does hairy conditional typedef
if test "$ac_cv_header_inttypes_h" != "yes"; then
XEmacs-Patches mailing list