Re: Request for rewiew: mbuffer-20050730

[Paul Howarth <paul city-fan org>]

Michael Schwendt wrote:
> On Wed, 03 Aug 2005 11:42:30 +0100, Paul Howarth wrote:
>
>
> > > > checking for library containing MD5_Init... no
> > > > checking for library containing mhash_init... no
> > > > checking for library containing MD5Init... no
> > > > configure: WARNING: could not find the mhash library containing the md5 
> > > > hash functions - md5 support disabled
> > >
> > >
> > > Detecting mhash_init would need mhash-devel BR, which is in Extras.
> >
> > The configure script doesn't look for mhash_init if it finds MD5_Init:
> >
> > ...
> > checking for library containing socket... none required
> > checking for library containing gethostbyname... none required
> > checking for library containing hstrerror... none required
> > checking for library containing MD5_Init... -lssl
> > configure: creating ./config.status
> > config.status: creating Makefile
> > config.status: creating mbuffer.1
> > config.status: creating config.h
> > ...
>
>
> So, to get MD5 support you really want to link against two OpenSSL libraries
> plus its Kerberos dependencies?

Yes, because the build won't be reproduceable otherwise. Systems with 
openssl-devel will use that implementation if they build the SRPM, 
whereas systems without it would get the mhash version.

This could be worked around by setting environment variable 
ac_cv_search_MD5_Init to "no" before running %configure so that MD5_Init 
wasn't found and it would then look for and find mhash_init. However, 
that would then result in a dependency on libmhash.so.2, which lives in 
%{_libdir} rather than /lib and hence breaks Alexander's aim of being 
able to use mbuffer with /usr not mounted.

Paul.