[Fedora-packaging] Is this allowed? Files and libs duplicated in subpackages
Stephen Gallagher
sgallagh at redhat.com
Wed Sep 4 20:21:56 UTC 2013
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On 09/04/2013 10:47 AM, Toshio Kuratomi wrote:
> On Tue, Sep 03, 2013 at 08:38:42AM -0400, Stephen Gallagher wrote:
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>>
>> On 09/01/2013 03:09 PM, Michael Schwendt wrote:
>>> https://bugzilla.redhat.com/1003196
>>>
>>> Based on this suspicious output
>>>
>>> mate-dictionary from mate-utils provides
>>> libmatedict.so.6()(64bit) mate-utils from mate-utils
>>> provides libmatedict.so.6()(64bit) required by:
>>> mate-dictionary-devel-1.6.0-7.fc20.x86_64 required by:
>>> mate-utils-devel-1.6.0-7.fc20.x86_64
>>>
>>> I've only verified in koji that lots of files are included in
>>> both sub-packages. Even the descriptions overlap.
>>>
>>> And there are even more subpackages, which only contain copies
>>> of files included in the base mate-utils package already. Why
>>> is that done? Why aren't RPM dependencies used to have the
>>> base-package depend on the multiple subpackages?
>>>
>>> So far, it has always been a packaging mistake to duplicate
>>> files (and their Provides as a consequence) in multiple
>>> subpackages.
>>
>>
>> Well, there are a few places where I can see duplicating files
>> making sense (but certainly not to the degree demonstrated in the
>> mate packages).
>>
>> For example, in the SSSD package, we duplicate the 'sssd_pac'
>> libexec binary in both the 'sssd-provider-ad' and
>> 'sssd-provider-ipa' plugin subpackages, rather than add useless
>> metadata for an extra common subpackage for both to depend on. It
>> seems wasteful to have a whole subpackage for one 150k binary.
>>
> https://fedoraproject.org/wiki/Packaging:Guidelines?rd=Packaging/Guidelines#DuplicateFiles
>
> So that would also be a packaging mistake. It's been many years
> since this was last touched though. IIRC, mschwendt raised the
> last issue with it so he may be best able to recall the
> justifications for this rule and whether the FPC should consider
> relaxing it.
>
For the record, I sent a patch to the SSSD upstream today to add a new
common sub-package for just that one file. It'll be cleaned up in the
next build.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.14 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
iEYEARECAAYFAlInlmQACgkQeiVVYja6o6PeBACfdhsGLWJmYiyysLY7SCzJfuzC
Y+sAn0m/csQLcFaPTYmO5ZFMXN1cNmUk
=bbM7
-----END PGP SIGNATURE-----
More information about the packaging
mailing list