[lfs-dev] glibc and rpc headers

Bruce Dubbs bruce.dubbs at gmail.com
Mon Aug 27 07:48:41 PDT 2012


Ken Moffat wrote:
> On Sun, Aug 26, 2012 at 09:30:56PM -0500, Bruce Dubbs wrote:
>> Starting a new thread.
>>
> [...]
>>
>> I'm starting to think that the problem is that we've built Chapter 6
>> glibc in 7.2 without the --enable-obsolete-rpc which would probably
>> solve the problem there.  For a 7.1 host, we'd need a note to install
>> the rpcnis-headers.tar.bz2 tarball as a part of Chapter 5 glibc.
>>
>> As in BLFS libtirpc:
>>
>> if [ ! r /usr/include/rpc/rpc.h ]; then
>>     tar -xvf ../rpcnis-headers.tar.bz2 -C /usr/include
>> fi
>>
>
>   Almost.  For 7.1, we didn't install them.  We fixed that a little
> while later, so my own 7.1 installs include them.  For 7.2 we _do_
> install them, at least if people follow all the instructions.  The
> switch is a cleaner way of doing it, but it installs a bit more
> (.x pascal headers - just like in the old versions) and is for our
> convenience going forward (one less thing to maintain).  But, adding
> that --enable-obsolete-rpc switch ought to be tested by another rc.
>
>> That's not something I want to do.  Perhaps the workaround of the sed
>> for Chapter 5 is the way to go.
>>
>   All versions of glibc before 2.14 installed these headers.  I would
> very much prefer that we require builders to install them if they
> are using an incomplete host system (probably LFS).

I think we have corner case here.  The only system that is a problem is 
7.1 without libtirpc being installed (or someone who didn't follow the 
book).  Now that I think of it, we could add to the host system 
requirements a check for /usr/include/rpc/rpc.h and if it doesn't exist 
just untar the glibc tarball and copy the header files from there to the 
host.  I doubt that they've changed and in that case they are not being 
used by anything else even if they have changed.

tar -xf   glibc-2.16.0.tar.xz
su -c 'cp glibc-2.16.0/sunrpc/rpc/*.h /usr/include/rpc'
rm -r     glibc-2.16.0

For this rare corner case, that's the minimum needed.

   -- Bruce



More information about the lfs-dev mailing list