[packagers] Setting up an rpmforge/rpmrepo dev environment

[Tom G. Christensen]

Dag Wieers wrote:
> On Fri, 30 Nov 2007, Tom G. Christensen wrote:
> 
>> Dag Wieers wrote:
>>> On Fri, 30 Nov 2007, Tom G. Christensen wrote:
>>>
>>>> Dag, if you're listening I'd strongly prefer rpmforge/rpmrepo to include
>>>> %{?dist} in the specfiles directly :)
>>> Dries and I am not in favor of this.
>>>
>> If you don't mind me asking, why not?
>>
>> In my view it makes it easier that you don't have to massage the specfile
>> before putting it through the buildsystem.
> 
> You should not care about this, as you do not need to care about what
> flags you send to rpmbuild, or whether you have to install dependencies
> yourself. All this should be done by the buildsystem.
>
> The fact that it is not done right now is not an excuse to not implement
> it or make it happen. The easy way is to add this %{?dist} and also
> include the repotag in this dist macro, but I am not in favor of that. The
> release indicates a number and the additional fields are mandatory and
> added by the buildsystem.
> 
> Besides, 2%{?dist} just looks ugly and makes it hard to see what the
> release is. It is useless.
> 
Thank your for explaining your position, it's given me something to 
think about.

<snip>
> Well, this is done to make it work on Fedora. As I said Matthias added
> these to build them as-is on Fedora (that does not know about the dtag).
> 
Yes I realize that.
It clashes with the set of macros I use in my buildroot though :)
I have a buildsys-build package that provides a set of macros that are 
run inside the buildroot, they determine all these values on the fly and 
only rely on the specfile to have %{?dtag: %{expand: %%define %dtag 1}} 
or similar.

With your switch from %dist to %dtag my life became much easier, perhaps 
I don't have to worry about the %fedora stuff, I'll figure it out.

> Checks like rhel > 4 make the SPEC file hard to understand and can lead to
> unexpected results. We prefer not to use it as such. If complete blocks
> are different it may be better to either keep the difference outside of
> the SPEC file or fork the SPEC file.
>
> Up till now that was never necessary though (and in fact we only fork a
> SPEC file to keep older versions for older distributions).
> 
I know your policy and I agree strongly with it.
The usage of %rhel and similar variables does not in my view negate this.
I don't use them much though, so perhaps you are right.

-tgc