Re: [Xen-devel] [PATCH v2] autoconf: check for Python.h header and -lpython* lib

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 20.03.12 at 14:06, "Jan Beulich" <JBeulich@xxxxxxxx> wrote:
>>>> On 12.03.12 at 14:35, Roger Pau MonnÃ<roger.pau@xxxxxxxxxxxxx> wrote:
>> 2012/3/12 Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>:
>>> Roger Pau Monne writes ("[PATCH v2] autoconf: check for Python.h header and 
>> -lpython* lib"):
>>>> autoconf: check for Python.h header and -lpython* lib
>>>
>>> Right, this is better.
>>>
>>>> +ac_previous_cppflags=$CPPFLAGS
>>> ...
>>>> +AC_CHECK_HEADER([Python.h], [],
>>>> +    [AC_MSG_ERROR([Unable to find Python development headers])],)
>>>> +AC_CHECK_LIB($python_lib, PyArg_ParseTuple, [],
>>>> +    [AC_MSG_ERROR([Unable to find a suitable python development 
>>>> library])])
>>>> +CPPFLAGS=$ac_previous_cppflags
>>>> +LDLFAGS=$ac_previous_ldflags
>>>
>>> But AFAICT we go to all this effort to find the right python flags and
>>> then we throw them away.  This is a bit odd.  Shouldn't we be passing
>>> this in some PYTHON_FOOFLAGS variables for the benefit of tools/python ?
>> 
>> This would only be needed to compile Python C modules, but since
>> python uses it's own standart build script (tools/python/setup.py) and
>> we already pass to the setup.py script the desired python version to
>> use, it already fetches the necessary includes/libs using it's own
>> mechanisms.
> 
> I'm sorry to say that, but this change is pretty incompatible: Not only
> does Python not provide any python-config prior to 2.5.x, but
> additionally was Ian's point above very valid: Stripping all the other
> -l options causes a configure failure for me:
> 
> configure:6176: checking for PyArg_ParseTuple in -lpython2.4
> configure:6201: gcc -o conftest  -g -O2   -g -O2  -I/usr/include/python2.4 
> -I/usr/include/python2.4 
> -L/usr/lib/python2.4/config conftest.c -lpython2.4   >&5
> /usr/lib/python2.4/config/libpython2.4.a(posixmodule.o): In function 
> `posix_tmpnam':
> (.text+0x81e): warning: the use of `tmpnam_r' is dangerous, better use 
> `mkstemp'
> /usr/lib/python2.4/config/libpython2.4.a(posixmodule.o): In function 
> `posix_tempnam':
> (.text+0x90c): warning: the use of `tempnam' is dangerous, better use 
> `mkstemp'
> /usr/lib/python2.4/config/libpython2.4.a(complexobject.o): In function 
> `complex_abs':
> (.text+0x54d): undefined reference to `hypot'
> /usr/lib/python2.4/config/libpython2.4.a(complexobject.o): In function 
> `_Py_c_pow':
> (.text+0x2418): undefined reference to `hypot'
> /usr/lib/python2.4/config/libpython2.4.a(complexobject.o): In function 
> `_Py_c_pow':
> (.text+0x242a): undefined reference to `pow'
> /usr/lib/python2.4/config/libpython2.4.a(complexobject.o): In function 
> `_Py_c_pow':
> (.text+0x243f): undefined reference to `atan2'
> /usr/lib/python2.4/config/libpython2.4.a(complexobject.o): In function 
> `_Py_c_pow':
> (.text+0x2470): undefined reference to `cos'
> /usr/lib/python2.4/config/libpython2.4.a(complexobject.o): In function 
> `_Py_c_pow':
> (.text+0x2483): undefined reference to `sin'
> /usr/lib/python2.4/config/libpython2.4.a(complexobject.o): In function 
> `_Py_c_pow':
> (.text+0x24f6): undefined reference to `exp'
> /usr/lib/python2.4/config/libpython2.4.a(complexobject.o): In function 
> `_Py_c_pow':
> (.text+0x2507): undefined reference to `log'
> /usr/lib/python2.4/config/libpython2.4.a(floatobject.o): In function 
> `float_rem':
> (.text+0x17c8): undefined reference to `fmod'
> /usr/lib/python2.4/config/libpython2.4.a(floatobject.o): In function 
> `float_divmod':
> (.text+0x25a8): undefined reference to `fmod'
> /usr/lib/python2.4/config/libpython2.4.a(floatobject.o): In function 
> `float_pow':
> (.text+0x296f): undefined reference to `pow'
> /usr/lib/python2.4/config/libpython2.4.a(floatobject.o): In function 
> `float_pow':
> (.text+0x2a03): undefined reference to `fmod'
> /usr/lib/python2.4/config/libpython2.4.a(dynload_shlib.o): In function 
> `_PyImport_GetDynLoadFunc':
> (.text+0x12e): undefined reference to `dlsym'
> /usr/lib/python2.4/config/libpython2.4.a(dynload_shlib.o): In function 
> `_PyImport_GetDynLoadFunc':
> (.text+0x195): undefined reference to `dlopen'
> /usr/lib/python2.4/config/libpython2.4.a(dynload_shlib.o): In function 
> `_PyImport_GetDynLoadFunc':
> (.text+0x1c3): undefined reference to `dlsym'
> /usr/lib/python2.4/config/libpython2.4.a(dynload_shlib.o): In function 
> `_PyImport_GetDynLoadFunc':
> (.text+0x232): undefined reference to `dlerror'
> /usr/lib/python2.4/config/libpython2.4.a(thread.o): In function 
> `PyThread_release_lock':
> (.text+0x49): undefined reference to `sem_post'
> /usr/lib/python2.4/config/libpython2.4.a(thread.o): In function 
> `PyThread_acquire_lock':
> (.text+0xc7): undefined reference to `sem_wait'
> /usr/lib/python2.4/config/libpython2.4.a(thread.o): In function 
> `PyThread_acquire_lock':
> (.text+0xe2): undefined reference to `sem_trywait'
> /usr/lib/python2.4/config/libpython2.4.a(thread.o): In function 
> `PyThread_free_lock':
> (.text+0x1f2): undefined reference to `sem_destroy'
> /usr/lib/python2.4/config/libpython2.4.a(thread.o): In function 
> `PyThread_allocate_lock':
> (.text+0x267): undefined reference to `sem_init'
> /usr/lib/python2.4/config/libpython2.4.a(thread.o): In function 
> `PyThread_start_new_thread':
> (.text+0x568): undefined reference to `pthread_create'
> /usr/lib/python2.4/config/libpython2.4.a(thread.o): In function 
> `PyThread_start_new_thread':
> (.text+0x586): undefined reference to `pthread_detach'
> /usr/lib/python2.4/config/libpython2.4.a(posixmodule.o): In function 
> `posix_forkpty':
> (.text+0x3189): undefined reference to `forkpty'
> /usr/lib/python2.4/config/libpython2.4.a(posixmodule.o): In function 
> `posix_openpty':
> (.text+0x3218): undefined reference to `openpty'
> collect2: ld returned 1 exit status
> 
> As you can see, all unresolved symbols revolve around symbols from
> those other libraries - the symbol looked for (PyArg_ParseTuple) is not
> among them.

A sketched out partial fix for this might be

--- a/tools/m4/python_devel.m4
+++ b/tools/m4/python_devel.m4
@@ -2,17 +2,18 @@ AC_DEFUN([AX_CHECK_PYTHON_DEVEL], [
 ac_previous_cppflags=$CPPFLAGS
 CPPFLAGS="$CFLAGS `$PYTHON-config --includes`"
 ac_previous_ldflags=$LDFLAGS
+python_other_libs=
 for flag in `$PYTHON-config --ldflags`
 do
     case $flag in
     -L*)
-        LDFLAGS="$LDLFAGS $flag"
+        LDFLAGS="$LDFLAGS $flag"
         ;;
     -lpython*)
         python_lib=`echo $flag | sed 's/^-l//'`
         ;;
     -l*)
-        # Ignore other libraries, we are only interested in testing python-dev
+        python_other_libs="$python_other_libs $flag"
         ;;
     *)
         AC_MSG_WARN([Strange ldflag found in $PYTHON-config output: $flag])
@@ -22,7 +23,8 @@ done
 AC_CHECK_HEADER([Python.h], [],
     [AC_MSG_ERROR([Unable to find Python development headers])],)
 AC_CHECK_LIB($python_lib, PyArg_ParseTuple, [],
-    [AC_MSG_ERROR([Unable to find a suitable python development library])])
+    [AC_MSG_ERROR([Unable to find a suitable python development library])],
+    [$python_other_libs])
 CPPFLAGS=$ac_previous_cppflags
 LDLFAGS=$ac_previous_ldflags
 ])

I didn't really re-generate tools/configure though (not sure what
autoconf version would be required, and surely this wouldn't match
the ones I have around), so can't claim this is syntactically correct.
For my purposes I instead patched tools/configure manually.

Partial because it doesn't address the lack of python-config on pre-2.5
Python (we require >= 2.3).

I also wonder whether the -lpython* case shouldn't be cumulative in
case there's more than one matching option.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel