Should redhat-release be versioned or unversioned?

[Bruno Wolff III]

On Wed, Dec 18, 2013 at 18:29:41 -0800,
   Adam Williamson <awilliam at redhat.com> wrote:
>On Wed, 2013-12-18 at 14:56 -0600, Bruno Wolff III wrote:
>> While looking at bug 1044675 I noticed that redhat-release is unversioned
>> in fedora-release and versioned in generic-release. I would expect it
>> to be the same in both of these packages. I think it probably makes the
>> most sense to version it for anything that is still using that, but wanted
>> to check if other people had good reasons for doing it one way or the
>> other.
>
>I've already looked into this. The only reason I didn't fix it already
>is that generic-release's tarball is secret sauce, there's no
>instructions on generating it in the spec file.

Lovely. OK won't rush into anything on this side then. (I might try to 
figure something out and consult with people to see what they think 
if I think I figured it out.)

>The bug I was looking at was
>https://bugzilla.redhat.com/show_bug.cgi?id=1040607 , where the
>redhat-release provide being versioned caused a problem (whereas in your
>case it was the other way around).

I think both cases are wrong now. I think yum used to do something 
smarter (or perhaps dumber) about grabbing the version. If I switch 
over to using system-release I'm going to hit the same problem. The 
problem I was seeing seems likely to have been triggered by the same 
yum update. So we'll probably want to make the same changes to 
system-release that we do to redhat-release.

>I agree it makes sense for them to match, to provide maximum similarity
>of behaviour, but we shouldn't version fedora-release's as
>{version}-{release} if we make them versioned, it appears! Either both
>unversioned, or both versioned just {version}, I think.

I assume you mean redhat-release above. fedora-release shouldn't be 
in a provides.

I'll also see if I can find some documentation on provides for how it 
is supposed to work regarding version and release.