Re: [lmi] using -std=c++0x instead of -std=c++98?

[Vaclav Slavik]

On Wed, 2010-03-17 at 19:32 +0000, Greg Chicares wrote:
> It's perfectly okay to require '-std=c++0x' instead of '-std=c++98':
> lmi doesn't need to be backward compatible with old versions of the
> C++ standard in this sense. 

Can you please apply this patch, then?


diff --git a/configure.ac b/configure.ac
index 9a63b62..58ac666 100644
--- a/configure.ac
+++ b/configure.ac
@@ -586,7 +586,7 @@ if test "x$GCC" == "xyes"; then
 fi
 
 if test "x$GXX" == "xyes"; then
-    CXXFLAGS="$CXXFLAGS $CFLAGS_gcc_common -std=c++98 \
+    CXXFLAGS="$CXXFLAGS $CFLAGS_gcc_common \
         -Wctor-dtor-privacy \
         -Wdeprecated \
         -Wnon-template-friend \
@@ -594,6 +594,15 @@ if test "x$GXX" == "xyes"; then
         -Wpmf-conversions \
         -Wsynth"
 
+    dnl g++ 4.4 doesn't compile wxWidgets code that relies on use of empty
+    dnl macro arguments, because it's behavior undefined by the standard
+    dnl prior to C++0x or C99; if the compiler is too old to understand
+    dnl -std=c++0x, it's also too old to complain about this and so we can
+    dnl use -std=c++98
+    AX_CXX_CHECK_FLAG([-std=c++0x],[],[],
+                      [CXXFLAGS="$CXXFLAGS -std=c++0x"],
+                      [CXXFLAGS="$CXXFLAGS -std=c++98"])
+
     dnl we need to use this option with g++ 4.3 to prevent its complaints
     dnl about "-funit-at-a-time is required for inlining of functions that are
     dnl only called once" in debug builds


I didn't do the equivalent for Windows build, because the newest GCC
version supported by it is 4.3 and that one didn't exhibit the problem
yet.

Thanks,
Vaclav