rpm installation tests / script tests

[Thorsten Leemhuis]

On 11.07.2007 16:40, Florian La Roche wrote:
> On Wed, Jul 11, 2007 at 03:10:45PM +0200, Thorsten Leemhuis wrote:
>> On 11.07.2007 14:51, Florian La Roche wrote:
>>> On Wed, Jul 11, 2007 at 07:41:55AM -0500, Rex Dieter wrote:
>>>> Florian La Roche wrote:
>>>>> anything holding back an update of the default snippets we
>>>>> propose for Fedora until a more permanent solution in
>>>>> rpm/tools is available?
>>>> The Snippets wiki has been updated, after getting feedback on the
>>>> fedora-desktop list.
>>> great news, very good.
>> +1, but will somebody kind of "enforce" this? Sure, it's no big
>> problems, but nevertheless would be nice to get fixed.
>>
>> How about a (script based?) change directly in cvs for all packages that
>> follow exactly what was in the "Snippets wiki" until now or was similar
>> to it?
>>
>> That would be just a bit of work for one person (yes, I'm volunteering
>> to do that if FESCo backs the idea; no, I'm not willing to write a
>> proposal for FESCo to discuss that, as that often is way more work then
>> actually doing the change), no boring work for packagers and the problem
>> will actually vanish then soon (¹).
> 
> +1 to get such things very fast into cvs and new packages built.

Well, it seems some people like tibbs dislike the idea with the "getting
packages built". I'm unsure myself what the proper way is to build or
not to build in devel -- normally I'd say "build", but for an issue like
this it might be acceptable to not build.

BTW, related question: should we commit the change not only to devel but
to the F-7 tree as well? Surely we should not build the packages
afterwards there, but if the packages get build sooner or later then the
fix will simply get it. And it avoids that people start to update a
package from the F-7 branch and copy the updated spec file to devel
afterwards and delete the change that was done in between.

> Usually
> most package owners are ok with such "packaging changes", but there is
> no official process for this until now.

I tend to agree with "most", but those that are not ok with such
"packaging changes" will yell loudly. So this for sure is something
where FESCo needs to come into the game.

>> Alternative would be to sun a spam-o-magic script that pokes people
>> regularly until they fix stuff is getting fixed.
> We all enjoy nag mails and huge amounts of new bugzilla requests.
> Maybe we'll also find smart ways with cvs commits across all packages
> at some point.

+1 -- we don't need a wiki style approach, but we IMHO should get a bit
more into the "wiki style" direction in the devel tree (as long as the
release is still some weeks away).

CU
knurd