[Fedora-packaging] Re: BuildRoot

[Ralf Corsepius]

On Tue, 2006-07-25 at 18:16 +0200, Axel Thimm wrote:
> On Tue, Jul 25, 2006 at 05:55:03PM +0200, Ralf Corsepius wrote:

> > - Doesn't work in your several %arch's case.
> 
> I didn't want to obfuscate it, better use sane and common defaults.
Well, your %buildroot isn't sane. It clashes and breaks on %arch

> The argument about arch was relative and not absolute anyway: "arch is
> more important than id, therefor if we skip arch, we need to skip id".
That's your argumentation. 

Mine is: The current build root supports id, but breaks on arch (==
defect of the recommendation) => We should fix this.

> But the scheme above even takes care of your multiuser-
> build-the-same-package-corner-case, so at least you have no reason not
> to be happy.
== no substantial progress on features in comparison to the current FE
recommendation.

> > - Do %name, %version, %release always expand correctly (Rpm suffers from
> >   a bug, where at least %name or %version (I don't recall exactly)
> >   occasionally is not being expanded correctly)?
> 
> URL?

Sorry, none.

>  I've never seen a macro fail using name/version and I use them
> quite a lot.

I once encountered it when simultaneously building several packages from
several source packages in one rpm.spec.

It had been a spec similar to this
Name: xxx
Version: 1
Release: 0
..
Source0: xxx-1.tar.gz
Source1: yyy-2.tar.gz
...
%package -n yyy
Name: yyy
Version: 2
..

In this case, the order of rpm sections (%build, %install etc.) is of
essential importance. Depending on where they are located, %version or 
%name expands to either xxx or yyy rsp. 1 or 2.

Unfortunately I don't have an example at hand to reproduce it.

Ralf