Re: Requires(...) vs. PreReq

[Jeff Johnson <n3npq nc rr com>]

Enrico Scholz wrote:

> Hello,
>
> on the question whether to use Requires(post,...) or PreReq, in
> https://bugzilla.fedora.us/show_bug.cgi?id=332#c40 it is said that
> Jeff Johnson said
>
> |   [...] and this is premature, you instally have syntax issues with
> |   older versions of rpm, as well as polluting rpm's context markers.
> |   The Right Thing To Do is to automagically generate scriptlet
> |   dependencies, [...]
>
> I know neither the context nor the age of this quote. Because the recent
> rpm documentation (the tsort-document) mentions the Requires(...) syntax,
> I am asking if
>
> | Requires(post,preun): /sbin/chkconfig
>
> or
>
> | PreReq: /sbin/chkconfig
>
> should be used with rpm>=4?
>  
Probably best atm is a simple
   Requires: /sbin/chkconfig

This has the same guarantee as PreReq:, without the context markers.

While the context markers are useful, particularly for breaking 
dependency loops precisely,
there are very few packages that need that degree of precision.

The bigger pain imho is breaking legacy builds by using new fangled 
syntax needlessly.
For example, the latest autoconf package uses the new-fangled syntax for 
no important
purpose, breaking builds on a package that is widely used and needed.

Retrofitting forward compatibility by releasing backward compatible versions
of rpm is the best fix I can figger, so I'm going to try to do 6 
releases of rpm in the
next year, legacy distros too.

73 de Jeff