Aggregation upstream projects are BAD (kdesdk for example)

Jon Ciesla limb at
Sat Sep 8 22:58:24 UTC 2007

> On Sat, 2007-09-08 at 12:09 -0400, Colin Walters wrote:
>> On 9/8/07, Hans de Goede <j.w.r.degoede at> wrote:
>>         But then I got a comment to the review it was already in
>>         Fedora in kdesdk. But
>>         then why on earth doesn't kdesdk have a Provides umbrello so
>>         that yum install
>>         umbrello works? Or even better an umbrello sub-package?
>> Better search would solve this problem, too:
>> (I'm not sure there's a hard line for when you basically have to give
>> up on categorization/naming standards and just rely on search, but I
>> think the Fedora package collection is past it)
> You don't want 'yum search' to always search every file in the
> distribution. Searching every file will just take a long time. And, in
> fact, you will routinely find that what you're searching for in the
> package is never explicitly listed in any of the file lists.
> so that leaves us two options:
>   - we include some sort of keyword tag in the pkg metadata
>   - we include some sort of keyword metadata file in the repodata
> Doing either of those and having yum search them if they are available
> is do-able. We talked about this before, iirc. We just need to decide
> which one is more palatable and do it.

Or, utilize pkgdb for this task.  Maybe capture the data at build time
with some sort of hook in koji?  A tiny delay in each build, while the
equivalent of an rpm -qa foo is dumped into pkgdg.  This could then make
not only searching but finding conflicts a snap.  You would even build in
a form where you could paste the rpm -qa foo of a new package during the
review step and it would check the pkgdb so you know with absolute
certainty that the new package does not conflict with something existing,
or if it does, you know where.

> -sv
> --
> fedora-devel-list mailing list
> fedora-devel-list at

novus ordo absurdum

More information about the fedora-devel-list mailing list