Re: [rpm-list] db version mismatch with a twist

[Jeff Johnson <n3npq jbj gmail com>]

On Nov 14, 2006, at 5:47 PM, Kevin Cosgrove wrote:

> A while ago I posted about building rpm 4.4.6 on both Solaris 2.8 and
> on RHEL 3, then trying to use the same rpm database for both.  After
> a bit of work to get all of the rpm and db stuff to link staticly for
> both systems the result is the same.  Jeff Johnson suggested that
> there might be a db issue with big endian and little endian involved.
> At that point I gave up and decided to split my database into one for
> Solaris and one for Linux.

Um, no.

I have personally read big-endian databases on little-endian i386 and  
vice versa.

I know of no issues wrto endianness and an entire rpmdb.

What I said is that mixing endianness in the same database has not  
been tested.

FYI, Berkeley DB supports mixed endian-ness, or at least appears to  
(I have not
checked whether the attribute is per insert or per-database, likely  
per-database).

The only hard part about attempting to support mixed endian records  
in rpm is
deciding which endianness to choose when installing, say, a noarch  
package,
or, say, installing an other-endian package for execution with a  
wrong-endian
emulator.

hth

73 de Jeff