On Jun 9, 2005, at 12:17 PM, Andreas Fester wrote:
Hi Tommi, Curt and Christophe,
Tommi, you once sent a patch for autoconf regarding the UTF8 and WCHAR
definitions. It has been partly applied some days ago, but the
definition of the macros is still missing from configure.in:
+AC_MSG_CHECKING([wchar_t])
+AC_COMPILE_IFELSE(AC_LANG_SOURCE(wchar_t* w;), [have_wchar_t=yes],
[have_wchar_t=no])
+AC_MSG_RESULT($have_wchar_t)
+
+if test "$have_wchar_t" = "yes"
+ then
+ AC_SUBST(LOGCHAR_IS_UTF8, 0)
+ AC_SUBST(HAS_WCHAR_T, 1)
+ else
+ AC_SUBST(LOGCHAR_IS_UTF8, 1)
+ AC_SUBST(HAS_WCHAR_T, 0)
+ fi
There was some discussion about this issue three to four weeks
ago about the dependency between LOGCHAR_IS_UTF8 and HAS_WCHAR_T.
Christophe, you talked about an improved patch - did you make any
progress yet? Did you intend to simply add a configure option
like --with-utf8 or is there some other "magic" which can
automatically
determine whether or not to use utf8?
I am no ant expert, but if I interpret the build.xml correctly
it sets
1) has.wchar_t to 0 if the platform supports wchar_t, to 1 otherwise
Vice-versa, has.wchar_t = 1 if platform supports wchar_t. If set 0,
no wchar_t API methods are compiled and mbstowcs et al cannot be used
for converting from default char encoding
2) utf-8 to 1 if has.wchar_t is 0 *or* utf-8 is explicitly requested
during configuration (would then be something like
"configure --with-utf8"),
otherwise utf-8 is set to 0.
Using utf-8 is likely the best default for 'nix. So I would
recommend a switch to indicate that wchar_t should be used as
logchar. Most 'nix should have wchar_t types but use UTF-8 as logchar.
