On Jan 20, 2006, at 12:30 PM, Tim Mooney wrote:
In regard to: AutoReqProv or AutoReq/AutoProv?, Dor-Shifer Amit
said (at...:
Hi.
My issue is this: an rpm is created in a dev. environment, where
all of
its dependencies are fulfilled, yet is installed in such environments
where those deps are only partially fulfilled. In order to
alleviate the
dep clash during installation, the maintainer builds the rpm with
"AutoReqProv=no". Yet this causes the RPM not to auto-generate any
"Provides" clauses. Packages depending on this rpm fail installing
because of this.
It is possible with modern/recent RPM to turn off dependency detection
by changing the appropriate macros. I've never tried to do this, but
my assumption is that you first turn off the internal dependency
generator, and then redefine the external dependency generator to
something that doesn't output any dependencies (something like
"exit 0").
The issue is a bit more complex than "turn off the internal
dependency generator".
Multilib packaging depends on files being identified with a ELF32/
ELF64 color.
In addition, internally generated dependencies are attached to files
so that the
dependency color can be computed from the attached file's ELF32/ELF64
color.
All the above mechanism is disabled when the internal dependency
generator
is disabled.
So it's impossible to build a package containing ELF files that can
be installed correctly on a
multilib system.
If you don't *need* to install on a multilib system, by all means,
generate dependencies
any way you wish, manually, with find-requires etc, or any other way
that you wish.
Just don't expect rpmlib to install your package on a multilib system
correctly.
Otherwise the dependency generation is entirely forward and backward
compatible
with deployed versions of rpm, equivalent to setting
%_transaction_color 0
Still, I'm not sure I understand the issue. If you generate an RPM
that says it has no dependencies (but it really does) and you install
it on some system that may or may not have those dependencies
installed,
are you expecting that it will work?
My question too.
73 de Jeff
_______________________________________________
Rpm-list mailing list
Rpm-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/rpm-list
