于 2011年02月18日 16:46, Jiri Denemark 写道:
On Fri, Feb 18, 2011 at 16:21:20 +0800, Osier Yang wrote:
> 于 2011年02月18日 11:00, Wen Congyang 写道:
>> +DEVMAPPER_CFLAGS=
>> +DEVMAPPER_LIBS=
>> +PKG_CHECK_MODULES([DEVMAPPER], [devmapper>= $DEVMAPPER_REQUIRED], [],
[DEVMAPPER_FOUND=no])
>> +if test "$DEVMAPPER_FOUND" = "no"; then
>> + # devmapper is missing pkg-config files in ubuntu, suse, etc
>> + save_LIBS="$LIBS"
>> + save_CFLAGS="$CFLAGS"
>> + DEVMAPPER_FOUND=yes
>> + AC_CHECK_HEADER([libdevmapper.h],,[DEVMAPPER_FOUND=no])
>> + AC_CHECK_LIB([devmapper], [dm_task_run],,[DEVMAPPER_FOUND=no])
>> + DEVMAPPER_LIBS="-ldevmapper"
>> + LIBS="$save_LIBS"
>> + CFLAGS="$save_CFLAGS"
>> +fi
>> +if test "$DEVMAPPER_FOUND" = "no" ; then
>> + AC_MSG_ERROR([You must install device-mapper-devel/libdevmapper>=
$DEVMAPPER_REQUIRED to compile libvirt])
>> fi
>> AC_SUBST([DEVMAPPER_CFLAGS])
>> AC_SUBST([DEVMAPPER_LIBS])
>
> Thanks for the fix, actually It can be reproduced simply by:
> % ./autogen.sh --with-storage-mpath=no
> % make
>
> As src/util/util.c invokes "dm_is_dm_major".
Yes it does, but I'm not convinced this is the right approach to fix it. Why
should libvirt always require device-mapper libraries? It only needs it for
disk and mpath storage drivers so if neither of them is built, we shouldn't
check for the library and we shouldn't call dm_is_dm_major.
Yes, it's better.
NACK to the current patch.
Jirka