| Message ID | 1455850625-9624-3-git-send-email-cardoe@cardoe.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Doug Goldstein writes ("[PATCH v2 3/4] m4/python: fix checks for Python library support"):
> AC_CHECK_LIB() was running gcc -Llib -lm -lutils conftest.c which on
> platforms that do as needed operations by default will result in
> underlinking. Instead AC_CHECK_LIB() suggests supplying the extra
> libraries necessary in a 5th argument.
I'm afraid I don't quite follow why the code in the patch has the
effect you desire. Why does setting LDFLAGS not work ? Forgive me if
I'm being dense - I'm afraid I have a cold.
Regards,
Ian.
On 2/22/16 10:22 AM, Ian Jackson wrote: > Doug Goldstein writes ("[PATCH v2 3/4] m4/python: fix checks for Python library support"): >> AC_CHECK_LIB() was running gcc -Llib -lm -lutils conftest.c which on >> platforms that do as needed operations by default will result in >> underlinking. Instead AC_CHECK_LIB() suggests supplying the extra >> libraries necessary in a 5th argument. > > I'm afraid I don't quite follow why the code in the patch has the > effect you desire. Why does setting LDFLAGS not work ? Forgive me if > I'm being dense - I'm afraid I have a cold. > > Regards, > Ian. > You need to put your libraries after the sources. e.g. gcc -W -march=native -Llibs -o program program.c -lm -lcustom Not before. That's why there's a difference between LIBS and LDFLAGS. The proper order is: COMPILER CFLAGS LDFLAGS SOURCES LIBS The reason is the compiler reduces the symbols to only the symbols used in each stage. By putting the libraries first the symbols are all reduced out and the result is a compiler failure. I can link you to the Gentoo docs on this: https://wiki.gentoo.org/wiki/Project:Quality_Assurance/As-needed#Failure_in_..2Fconfigure
On 2/22/16 10:22 AM, Ian Jackson wrote: > Doug Goldstein writes ("[PATCH v2 3/4] m4/python: fix checks for Python library support"): >> AC_CHECK_LIB() was running gcc -Llib -lm -lutils conftest.c which on >> platforms that do as needed operations by default will result in >> underlinking. Instead AC_CHECK_LIB() suggests supplying the extra >> libraries necessary in a 5th argument. > > I'm afraid I don't quite follow why the code in the patch has the > effect you desire. Why does setting LDFLAGS not work ? Forgive me if > I'm being dense - I'm afraid I have a cold. > > Regards, > Ian. > In fact this pattern is repeated all over the code base in makefiles: libxenctrl.so.$(MAJOR).$(MINOR): $(CTRL_PIC_OBJS) $(CC) $(LDFLAGS) $(PTHREAD_LDFLAGS) -Wl,$(SONAME_LDFLAG) -Wl,libxenctrl.so.$(MAJOR) $(SHLIB_LDFLAGS) -o $@ $^ $(LDLIBS_libxentoollog) $(LDLIBS_libxenevtchn) $(LDLIBS_libxengnttab) $(LDLIBS_libxengntshr) $(LDLIBS_libxencall) $(LDLIBS_libxenforeignmemory) $(PTHREAD_LIBS) $(APPEND_LDFLAGS) COMPILER CFLAGS/LDFLAGS -o TARGET SOURCE LIBS
On 2/22/16 1:14 PM, Doug Goldstein wrote: > On 2/22/16 10:22 AM, Ian Jackson wrote: >> Doug Goldstein writes ("[PATCH v2 3/4] m4/python: fix checks for Python library support"): >>> AC_CHECK_LIB() was running gcc -Llib -lm -lutils conftest.c which on >>> platforms that do as needed operations by default will result in >>> underlinking. Instead AC_CHECK_LIB() suggests supplying the extra >>> libraries necessary in a 5th argument. >> >> I'm afraid I don't quite follow why the code in the patch has the >> effect you desire. Why does setting LDFLAGS not work ? Forgive me if >> I'm being dense - I'm afraid I have a cold. >> >> Regards, >> Ian. >> > > You need to put your libraries after the sources. e.g. > > gcc -W -march=native -Llibs -o program program.c -lm -lcustom > > Not before. That's why there's a difference between LIBS and LDFLAGS. > The proper order is: > > COMPILER CFLAGS LDFLAGS SOURCES LIBS > > The reason is the compiler reduces the symbols to only the symbols used > in each stage. By putting the libraries first the symbols are all > reduced out and the result is a compiler failure. > > I can link you to the Gentoo docs on this: > https://wiki.gentoo.org/wiki/Project:Quality_Assurance/As-needed#Failure_in_..2Fconfigure > > Does this answer your question?
Doug Goldstein writes ("Re: [PATCH v2 3/4] m4/python: fix checks for Python library support"):
> You need to put your libraries after the sources. e.g.
> gcc -W -march=native -Llibs -o program program.c -lm -lcustom
Of course. I really was quite thick last week! I have queued this
patch now.
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Ian.
diff --git a/m4/python_devel.m4 b/m4/python_devel.m4 index deff19e..05ea4ef 100644 --- a/m4/python_devel.m4 +++ b/m4/python_devel.m4 @@ -10,9 +10,9 @@ AS_IF([test x"$pyconfig" = x"no"], [ print "-I" + distutils.sysconfig.get_config_var("INCLUDEPY")'`" CPPFLAGS="$CPPFLAGS `$PYTHON -c 'import distutils.sysconfig; \ print distutils.sysconfig.get_config_var("CFLAGS")'`" - LDFLAGS="$LDFLAGS `$PYTHON -c 'import distutils.sysconfig; \ + PYTHON_LIBS="$LDFLAGS `$PYTHON -c 'import distutils.sysconfig; \ print distutils.sysconfig.get_config_var("LIBS")'`" - LDFLAGS="$LDFLAGS `$PYTHON -c 'import distutils.sysconfig; \ + PYTHON_LIBS="$LDFLAGS `$PYTHON -c 'import distutils.sysconfig; \ print distutils.sysconfig.get_config_var("SYSLIBS")'`" LDFLAGS="$LDFLAGS `$PYTHON -c 'import distutils.sysconfig; \ print "-L" + distutils.sysconfig.get_python_lib(plat_specific=1,\ @@ -25,12 +25,14 @@ AS_IF([test x"$pyconfig" = x"no"], [ dnl If python-config is found use it CPPFLAGS="$CFLAGS `$PYTHON-config --cflags`" LDFLAGS="$LDFLAGS `$PYTHON-config --ldflags`" + PYTHON_LIBS="$LIBS `$PYTHON-config --libs`" ]) AC_CHECK_HEADER([Python.h], [], [AC_MSG_ERROR([Unable to find Python development headers])],) AC_CHECK_LIB(python$ac_python_version, PyArg_ParseTuple, [], - [AC_MSG_ERROR([Unable to find a suitable python development library])]) + [AC_MSG_ERROR([Unable to find a suitable python development library])], + [$PYTHON_LIBS]) CPPFLAGS=$ac_previous_cppflags LDFLAGS=$ac_previous_ldflags ])
AC_CHECK_LIB() was running gcc -Llib -lm -lutils conftest.c which on platforms that do as needed operations by default will result in underlinking. Instead AC_CHECK_LIB() suggests supplying the extra libraries necessary in a 5th argument. Signed-off-by: Doug Goldstein <cardoe@cardoe.com> --- CC: Ian Jackson <ian.jackson@eu.citrix.com> CC: Stefano Stabellini <stefano.stabellini@eu.citrix.com> CC: Ian Campbell <ian.campbell@citrix.com> CC: Wei Liu <wei.liu2@citrix.com> change since v1: - don't screw up the white space --- m4/python_devel.m4 | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)