Package EVR problems in FC+FE 2007-03-22

[Michael Schwendt]

On Fri, 23 Mar 2007 19:59:32 +0100, Axel Thimm wrote:

> On Fri, Mar 23, 2007 at 07:40:51PM +0100, Michael Schwendt wrote:
> > On Fri, 23 Mar 2007 19:22:15 +0100, Axel Thimm wrote:
> > 
> > > On Fri, Mar 23, 2007 at 11:23:34AM +0100, Michael Schwendt wrote:
> > > > On Thu, 22 Mar 2007 20:34:02 -0400 (EDT), buildsys wrote:
> > > > 
> > > > > limb AT jcomserv.net:
> > > > >     ettercap
> > > > >       FE5 > FE6 (0:0.7.3-13.3.fc5 > 0:0.7.3-12.fc6)
> > > > >       FE5 > FE7 (0:0.7.3-13.3.fc5 > 0:0.7.3-12.fc7)
> > > > >     
> > > > >     gnubg
> > > > >       FE5 > FE7 (0:20061119-8.fc5.2 > 0:20061119-7.fc7)
> > > > >       FE6 > FE7 (0:20061119-8.fc6 > 0:20061119-7.fc7)
> > > > >     
> > > > >     xmoto-edit
> > > > >       FE5 > FE6 (0:0.2.4-6.1.fc5 > 0:0.2.4-6.fc6)
> > > > >       FE5 > FE7 (0:0.2.4-6.1.fc5 > 0:0.2.4-6.fc7)
> > > > >     
> > > > > miker5slow AT grandecom.net:
> > > > >     etherape
> > > > >       FE5 > FE6 (0:0.9.7-4.1.fc5 > 0:0.9.7-4.fc6)
> > > > >       FE5 > FE7 (0:0.9.7-4.1.fc5 > 0:0.9.7-4.fc7)
> > > > 
> > > > The issues with "ettercap", "xmoto-edit" and "etherape" are one of the
> > > > infamous problems when adding a %dist tag.
> > > 
> > > Ehem. All of the above have the same inequality relation if you drop
> > > the disttag ...
> > 
> > Impossible, because then you cannot tag and build the packages in the way
> > they have been built above. Only due to the added %disttag, it was
> > possible to keep the rest of %release equal across multiple branches.
> 
> You're setting 13.3 and 12 as well as 4.1 and 4 to be equal. OK, that
> must count as cheating at the very least. ;)

Haha, almost funny.

So, you've taken the bait and continued with your very own way of
thinking. What do you want to prove this time? ;)

That Joe Packager can do the following?

       FE5 > FE7 (0:0.7.3-100 > 0:0.7.3-12)

Yes, look closely. In a no-dist-tag scheme, packager can break versioning
and upgrade paths in various ways just as when using dist tags. What is
the point? Aiming at creating an endless thread once more?

The key difference is, that _normally_ without a dist tag the packager is
fully responsible for keeping release of the latest branch higher than for
old branches. Once that is achieved, e.g.

       FE5 < FE7 (0:0.7.3-5 < 0:0.7.3-7)

the packager can apply minor release bumps without any risk of breaking
the upgrade path and without comparing release digits and dist tags:

       FE5 < FE7 (0:0.7.3-5.1 < 0:0.7.3-7)
       FE5 < FE7 (0:0.7.3-5.2 < 0:0.7.3-7)
       ...
       FE5 < FE7 (0:0.7.3-5.20 < 0:0.7.3-7)

In my original post I didn't argue that a careful packager can't keep the
upgrade path sane when using a dist tag, e.g. by adding something at the
very right of %release. I pointed out a common mistake that is being done
when using dist tags. It's just a bigger pitfall, because the visible
part of %release is _equal_ for multiple branches.

And please, don't point out that the guidelines cover minor release
bumps with dist tags. That has been done before.

This will be my last reply in this thread.