HAVE_TIMEZONE test failing for a target

Tony Kimball alk@pobox.com
Wed Apr 25 22:15:00 GMT 2001


Just FYI:

libjava/configure.in:733:
   AC_TRY_COMPILE([#include <time.h>], [long z2 = timezone;],
     [AC_DEFINE(HAVE_TIMEZONE)
      AC_MSG_RESULT(yes)],
     [AC_MSG_RESULT(no)])])

Evidently the AC_TRY_COMPILE arguments are not matching the arguments
used to compile java/lang/natSystem.cc, because this is coming out
'yes', and HAVE_TIMEZONE is defined 1 for a target without the global
variable.  If this is true for one platform, it's probably true for
all platforms, i.e. the equivalence of the command lines in the two
cases (config test and natSystem.cc) is not assured, as it should be.

My case is: host=i686-pc-linux-gnu; target=i386-cygwin.  HEAD revision
of :pserver:anoncvs@gcc.gnu.org:/cvs/gcc.

The conflicting definition of timezone as char *timezone() is found in
$(prefix)/i386-cygwin/include/time.h. 

Me, I'll work around it.  Because I think autoconf is a bad prank.





More information about the Java mailing list