[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: bug relating to func_infer_tag
From: |
Bruno Haible |
Subject: |
Re: bug relating to func_infer_tag |
Date: |
Sun, 2 May 2010 16:01:39 +0200 |
User-agent: |
KMail/1.9.9 |
Hello Ralf,
> I think things would work if you used AC_PROG_CXX instead of
> hand-written similar functionality for setting CXX. Automake and
> Libtool specifically discover this macro as a sign for use of C++
This is exactly the problem. It looks for AC_PROG_CXX and nothing else.
It does not give me the freedom to use an alternative for AC_PROG_CXX.
In my case, I don't want to use AC_PROG_CXX because it aborts the entire
configure script if it does not find a C++ compiler. Other people may
want to use a different macro is they prefer some vendor compiler instead
of g++.
> may tack code onto the end of it.
Automake and Libtool modify the code of AC_PROG_CXX. Can't you come up
with a better design than this?
> Using AC_PROG_CXX should enable you to drop use of internal
> _AM_DEPENDENCIES([CXX]).
No, this is not an option for me.
Look at this: Automake apparently defines
CXXLINK = $(LIBTOOL) $(AM_LIBTOOLFLAGS) ...
by default, and
CXXLINK = $(LIBTOOL) --tag=CXX $(AM_LIBTOOLFLAGS) ...
if it has spotted an LT_LANG([CXX]) invocation.
Why does is not define
CXXLINK = $(LIBTOOL) --tag=CXX $(AM_LIBTOOLFLAGS) ...
unconditionally? When defining the variable CXXLINK, Automake knows that
its use it to link executables which contain C++ object files. That
requires the use of the C++ compiler, always. And the use of the C++
compiler as a libtool argument requires --tag=CXX. No?
Bruno