Re: [PATCH 2/2] libselinux: do not use relative path when creating libselinux symlinks

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 09/15/11 08:53, Guido Trentalancia wrote:
> On Thu, 2011-09-15 at 08:22 -0400, Christopher J. PeBenito wrote:
>> On 09/15/11 08:16, Guido Trentalancia wrote:
>>> On Thu, 2011-09-15 at 08:01 -0400, Christopher J. PeBenito wrote:
>>>> On 09/14/11 16:21, Eric Paris wrote:
>>>>> On Wed, 2011-09-14 at 15:18 -0400, Stephen Smalley wrote:
>>>>>> On Wed, 2011-09-14 at 14:50 -0400, Eric Paris wrote:
>>>>>>> At the moment we create a symlink:
>>>>>>>
>>>>>>> /usr/lib/libselinux.so -> ../../lib/libselinux.so.1
>>>>>>>
>>>>>>> This works if (and only if) $SHLIBDIR and $LIBDIR are different only by
>>>>>>> ../../.  Instead create a symlink from
>>>>>>>
>>>>>>> $LIBDIR/libselinux.so->$SHLIBDIR/libselinux.so.1
>>>>>>>
>>>>>>> Thus it works no matter what values one might use for LIBDIR and
>>>>>>> SHLIBDIR.
>>>>>>
>>>>>> I'm not sure this works the way you would want.  Consider rpm build of
>>>>>> libselinux - it does:
>>>>>> make DESTDIR="%{buildroot}" LIBDIR="%{buildroot}%{_libdir}"
>>>>>> SHLIBDIR="%{buildroot}/%{_lib}" BINDIR="%{buildroot}%{_sbindir}" install
>>>>>>
>>>>>> And then rpm collects up the files into the package.
>>>>>> But if the symlink encodes the full pathname used at make install time,
>>>>>> then it will be wrong on the final system when the rpm is installed.
>>>>>> Haven't actually tested that theory, but I think it is true.  Welcome to
>>>>>> hell.
>>>>>
>>>>> error: Symlink points to BuildRoot: /usr/lib64/libselinux.so
>>>>> -> /root/rpmbuild/BUILDROOT/libselinux-2.1.5-4.fc16.1.eparis.x86_64/lib64/libselinux.so.1
>>>>>
>>>>> GRRRRRR.   Every other package I see with similar symlinks seems to be
>>>>> an autoconf package and I can't understand how they work.  What we have
>>>>> isn't right, but I don't know how to fix it....
>>>>
>>>> Its easy.  The makefile needs to be tweaked so you don't have to specify %{buildroot} as part of the LIBDIR, SHLIBDIR, and BINDIR variables.  Then when you use those variables, you prepend DESTDIR as necessary.  So if you have usages like
>>>>
>>>> install foo $(BINDIR)
>>>>
>>>> it turns into
>>>>
>>>> install foo $(DESTDIR)(BINDIR)
>>>>
>>>> Then when you have your symlink you can do
>>>>
>>>> ln -s $(SHLIBDIR)/bar.so.1 $(DESTDIR)$(SHLIBDIR)/bar.so
>>>
>>> Yes, this is even better, although $(DESTDIR) is not necessary in the
>>> target, as SHLIBDIR already includes (begins with) DESTDIR:
>>>
>>> ln -sf $(SHLIBDIR)/$(LIBSO) $(SHLIBDIR)/$(TARGET)
>>
>> I think you're missing the point.  I said you need to stop specifying $(DESTDIR) as part of $(SHLIBDIR).  Otherwise you get the broken symlinks that Eric talks about above.
> 
> In any case, I think it then probably ends up taking more characters
> than using pushd/popd.

Perhaps it does.  I'd argue that its much clearer, thus more maintainable than pushd/popd.

-- 
Chris PeBenito
Tresys Technology, LLC
www.tresys.com | oss.tresys.com

--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to majordomo@xxxxxxxxxxxxx with
the words "unsubscribe selinux" without quotes as the message.


[Index of Archives]     [Selinux Refpolicy]     [Linux SGX]     [Fedora Users]     [Fedora Desktop]     [Yosemite Photos]     [Yosemite Camping]     [Yosemite Campsites]     [KDE Users]     [Gnome Users]

  Powered by Linux