[Gluster-devel] libgfapi.so rpm location?

Kaleb S. KEITHLEY kkeithle at redhat.com
Mon Feb 17 22:20:44 UTC 2014

On 02/17/2014 05:07 PM, Justin Clift wrote:
> Hi Kaleb,
> Just noticed a small weirdness with the GlusterFS rpms generated from
> from git master head.
> In glusterfs-api rpm:
>   * /usr/lib64/libgfapi.so.6
>   * /usr/lib64/libgfapi.so.6.0.0
> In glusterfs-api-devel rpm:
>   * /usr/lib64/libgfapi.so
> They should all be in the glusterfs-api rpm shouldn't they?

Nope. This is ELF sharedlib juju.

At compile time the linker looks for libfoo.so. It reads the SO_NAME, 
which is libfoo.so.x (or libfoo.so.x.y.z), from libfoo.so (which is just 
a symlink to libfoo.so.x); Later, at run-time, the dyld loads libfoo.so.x.

RPM's packaging rules say that libfoo.so.x (and libfoo.so.x.y.z) belong 
in the foo.rpm, but the libfoo.so symlink belongs in the foo-devel.rpm.

IOW a run-time only install only needs libfoo.so.x or libfoo.so.x.y.z.

A developer who wants to link with the lib needs the -devel rpm with 

The purist packagers like this because it keeps the file system from 
being filled with lib*.so symlinks that aren't needed except by developers.

Debian/Ubuntu-style dpkg has similar rules for their -dev dpkgs.

> If so, I can include the fix in my Glupy rename patch, as it's
> already moving a few things around for the API rpms.
> Regards and best wishes,
> Justin Clift
> --
> Open Source and Standards @ Red Hat
> twitter.com/realjustinclift



More information about the Gluster-devel mailing list