RE: dynamic dependency check

[Reznic Valery <Valery.Reznic@xxxxxxxxxxxx>]

Title: Message

 

-----Original Message-----
From: Dhanvi K [mailto:dhanvik@xxxxxxx]
Sent: Wednesday, January 21, 2004 2:34 PM
To: rpm-list@xxxxxxxxxx
Subject: Re: dynamic dependency check

Hi Scot

     I would like my spec file to dynamically pick up rpm requirements
depending on the OS on which the RPM is being installed. for e.g. If the
RPM is being installed on Fedora box, then the spec should pick up one
set of dependency rpms, while if its installing on a RedHat Enterprise
box,  then it should pick up another set of RPMS. 

RPM spec exist on the build computer, so spec file can't 

"dynamically pick up rpm requirements"

You can conditionally put one or another package as your rpm requirement.

But do you really need it ?

In most case rpm depend not on another rpm but on lib, interpreter, etc.

These kind of dependencies automatically generated during rpm build time.

In this case all you need is just rebuild your source rpm on each of the target

platforms

Valery

 

     I would like to know whether and how this is possible in the spec
file ? Thanks for your time and advice
    

Yes this is possible, though it requires that the OS variation be declared in 
some standard format by the OS...it is "often" found in /etc/. Though there 
is no standard format. One could use a structure to discover various points 
of interest on the operating system, but discovering with something is 
fedora, redhat, debian, slackware, and their various versions. And then one 
needs to consider user modified systems.
  

Yeah I realize that detecting various *NX flavors would be difficult job. My current project
requirements clearly specify 3 linux flavors ( fedora, mandrake , Redhat Enterprise ) on which we are going to support our product,
hence this would a pretty straight forward task.. I presume

Generally speaking...and I do say generally...the most important things to 
know are the kernel type and version, and libc type and version. All the 
other dependencies are really far too many to try and track/monitor with a 
spec file. The best thing to do is to make your spec file look for the 
various "differences" in pieces.  Use the %ifarch and %ifos directives. You

I will check out these directives.

 
can also use shell scriptlets for a great deal of your logic system. But 
again you are looking at a lot of work. One of the best soiurces of 
inspiration and consequentially also dispair are teh xfree86 spec files that 
mharris has developed. They cover a lot of system detection ground and may be 
of some help.
  

I will certainly check out the spec file specified.. Thanks for the patient reply.

But be ready for the fact that you aren't really going to be able to cover 
ever single instance without developing and maintaining the one single spec 
file becoming a full time job as you try to keep up with all the changing 
systems. Worry instead about covering standards and creating a standards 
compliant spec file. Allow users to "customize" the spec files for their 
needs instead. Keeping things simple will make things easier on your users, 
trying to cover every possibility will make things harder for you, and will 
also make things harder for your users when they DO find they must customize 
your spec file to work for their system.  Not everyone is system builder, and 
not everyone has the patience to dig through a complex logic structure.

Scot

  

Dhanvi

_______________________________________________ Rpm-list mailing list Rpm-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/rpm-list