On Tuesday 04 November 2014 16:11:39 Richard W.M. Jones wrote:
On Tue, Nov 04, 2014 at 04:35:30PM +0100, Pino Toscano wrote:
> Instead of querying distutils.sysconfig for the library name, just
> use the $PYTHON_LIBS returned by pkg-config. This makes sure the
> tests can link even with a libpython in a non-standard library
> directory. ---
>
> configure.ac | 10 ++--------
> 1 file changed, 2 insertions(+), 8 deletions(-)
>
> diff --git a/configure.ac b/configure.ac
> index 87e066c..1bf291c 100644
> --- a/configure.ac
> +++ b/configure.ac
> @@ -1216,20 +1216,14 @@ AS_IF([test "x$enable_python" !=
"xno"],[
>
> AC_MSG_RESULT([$PYTHON_EXT_SUFFIX])
>
> dnl Look for some optional symbols in libpython.
>
> - old_LIBS="$LIBS"
> -
> - PYTHON_BLDLIBRARY=`$PYTHON -c "import distutils.sysconfig;
> \
> - print
> (distutils.sysconfig.get_config_var('BLDLIBRARY'))"`>
> AC_CHECK_LIB([c],[PyCapsule_New],
>
> [AC_DEFINE([HAVE_PYCAPSULE_NEW],1,
>
> [Found PyCapsule_New in
> libpython.])],
>
> - [],[$PYTHON_BLDLIBRARY])
> + [],[$PYTHON_LIBS])
>
> AC_CHECK_LIB([c],[PyString_AsString],
>
> [AC_DEFINE([HAVE_PYSTRING_ASSTRING],1,
>
> [Found PyString_AsString in
> libpython.])],
>
> - [],[$PYTHON_BLDLIBRARY])
> -
> - LIBS="$old_LIBS"
> + [],[$PYTHON_LIBS])
>
> fi
>
> AC_SUBST(PYTHON_PREFIX)
Will this break hillu's change (commit ad5b5f96da53)?
I found that commit, yes (I was wondering why AC_CHECK_LIB with "c" as
primary library).
My commit should not (at least, it gave no changes on Fedora) break
things; it just uses the more reliable variable instead of a config var
which (at least to me) seems to be used only internally in python.
I can ask about it, anyway.
--
Pino Toscano