Re: Re: [rpm-list] %ifos

["Jeff Johnson" <n3npq jbj gmail com>]

On 11/15/06, Kevin Cosgrove <kevinc doink com> wrote:
...
> The cpu-vendor-os stuff seems like a much simpler route and I would
> be one to welcome it.
>
> I'm not too sure how that relates to my situation tho.  Here's what
> I'm doing:
>
> %build
> [some stuff]
> %ifos solaris2.8
>   %{_bindir}/egd.pl --kill checks/entropy
> %endif

The issue "relates" to what you are doing because

1) the "OS" is set by /etc/rpm/platform, otherwise uname(2) and a
string transorm

2) there are OS compatibility tables that are driven by the OS from 1) that
will determine whether "solaris2.8" is compatible.

No matter what, %ifos is handled during parsing on the build machine, not
from the OS on the install machine. If that's not what you want, then use
shell logic, not %ifos, in scriptlets.

> The above doesn't run the perl script on solaris2.8 machines at the
> time that I build the rpm packages.  "%ifos solaris2.8" outside one
> of the build stages, i.e. at the top of a spec file where variables
> are typically set, _does_ behave as if the os is solaris2.8.  For the
> linux os, %ifos works everywhere.  Is this something that can be
> altered by %_sysconfdir/rpm/platform contents?

rpm --showrc displays both build and insall arch/os, as well as the
compatibility
tables.

> My workaround is to do this near the top of the spec file
>
> %ifos solaris2.8
>   %define myscript %{_bindir}/egd.pl --kill checks/entropy
> %else
>   %define myscript echo do nothing
> %endif

All depends on whether you want build or install system logic. Macros
ars almost always expanded during build, not anywhere else.

hth

73 de Jeff