[NTLUG:Discuss] Rebuilding (G)libc from an SRPM

George Lass George.Lass at osc.com
Thu Sep 23 22:25:35 CDT 2004


Just FYI we are trying to rebuild libc to reduce the
standard timeout parameter for mounting NFS file systems
when the remote host is down. We want to reduce the
timeout to just a few seconds during startup.

George



-----Original Message-----
From: discuss-bounces at ntlug.org on behalf of Greg Camp
Sent: Thu 9/23/2004 8:54 PM
To: NTLUG Discussion List
Subject: RE: [NTLUG:Discuss] Rebuilding (G)libc from an SRPM
 
> On Thu, 2004-09-23 at 13:04, George Lass wrote:
> > I am attempting rebuild libc (or maybe glibc) for a redhat
> > 8.0 system. The originally installed libc in /lib has the
> > following characteristics:
> > -rwxr-xr-x    2 root     root      1327065 Sep  5  2002 
> libc-2.2.93.so
> > lrwxrwxrwx    1 root     root           14 Apr 27 05:51 libc.so.6 ->
> > libc-2.2.93.so
> > but when rebuilt (using rpm & rpmbuild) from 
> glibc-2.2.93-5.src.rpm it
> > looks like this:
> > -rwxr-xr-x    1 root     root     11255857 Sep 20 17:56 libc.so
> > lrwxrwxrwx    1 root     root            7 Sep 20 17:30 libc.so.6 ->
> > libc.so
> > I'm somewhat reluctant to actually install this thing given 
> that it is
> > nearly 10X the size of the original.  Anyone have any ideas?
> 
> Any reason why you rebuilt GLibC?
> Optimized for Athlon?
> Other?
> 
> I've rebuilt GLibC to optimize for Athlon.
> But I've never had any issues doing such.
> No size issues.
> 
> 11MB seems big for even a non-stripped GLibC.
> Possibly built with debugging or other symbols?
> 
> What parameters (to ./configure, or via "--target=") did you pass?
> 

We built it using the glibc.spec file:
rpmbuild -bc ./SPECS/glibc.spec

Should it be rebuilt with additional/different parameters?  Our thought
was to build it using the same parameters RedHat did originally, but
perhaps they aren't embedded in the .spec file as we had hoped.

Greg


_______________________________________________
https://ntlug.org/mailman/listinfo/discuss




More information about the Discuss mailing list