[Pkg-puppet-devel] puppet-agent_7.16.0-1_amd64.changes ACCEPTED into experimental, experimental

[Antoine Beaupré]

Hi zigo!

First, thanks for your hard work on the Puppet packages (and modules)!

On 2022-05-16 15:23:40, Thomas Goirand wrote:
> Hi Jerome,
>
> Thanks for caring about Puppet and for your quick reply.
>
> On 5/16/22 14:35, Jerome Charaoui wrote:
>> Le 2022-05-16 à 03 h 56, Thomas Goirand a écrit :
>>> On 5/15/22 12:00, Debian FTP Masters wrote:
>>>>     * Renamed service to "puppet-agent" with "puppet" alias
>>>>     * debian/control:
>>>>       * added new dependencies: ruby-concurrent and ruby-semantic-puppet
>>>>       * do not build puppet-master or puppet-common packages
>>>>       * build puppet -> puppet-agent transitional package
>>>>       * add Breaks/Replaces older puppet package
>>>
>>>
>>> Hi,
>>>
>>> We already did a transition puppet-agent -> puppet in the past, I 
>>> don't think we want to do that work on the opposite direction a few 
>>> release later.
>>>
>>> Jerome, could you please revert this, and keep the name puppet-agent 
>>> for the binary? It's ok if you add a Provides: if you like though.
>> 
>> I think we should definitely keep the source package named 
>> "puppet-agent" because this one only builds the agent whereas the old 
>> "puppet" 5.5 source package built both agent and server components.
>
> There's on consequence switching the source package name, and never the 
> need to ask anyone about it, so I honestly don't care about this.

What do you mean by "on consequence"? Did you mean "no consequence"?

>> As for the binary package I don't have a strong opinion either way, 
>> though its true that keeping the name "puppet" for the binary of puppet 
>> agent would at least avoid the need to update modules that manage the 
>> agent packages on Debian & family.
>
> I do (have a strong opinion). See the list of reverse dependency below. 
> If you wish to make that change backward (ie: back from puppet to 
> puppet-agent like it was 3 or 4 Debian releases ago), please also NMU 
> changes to all of the below packages. That's more than 90 packages to 
> change. All of that for almost no benefits (and no motivation given). If 
> you do have a real reason to do it, I'm fine with discussing it in this 
> list, but otherwise, I'm re-iterating: could you please reverse the 
> binary package name change?

But do those packages all really depend on the agent though?

It seems to me it would actually be a good thing to review those
dependencies, since we *are* splitting the server and agent part. All
the puppet-module-* packages, for example, probably do not *need* to
Depend on the puppet-agent package, that might just be a "Recommends",
right?

I otherwise happen to agree with lavamind's change: I think puppet-agent
is a better name for the binary package. It states clearly that it's
*just* the agent (whereas the "puppet" package before was possibly
shipping more, and had intermingled deps with the server). When we did
the previous switch, wasn't it exactly because the codebase was shared,
something that was changed upstream?

I wasn't around the last time the switch was made though, so maybe I
misunderstand that part of the history...

a.

-- 
Men are taught to apologize for their weaknesses, women for their
strengths.
                        - Lois Wyse