[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [scl.org] Component name in non-RH dist-gits

On 02/17/2015 09:21 AM, Remi Collet wrote:
Le 17/02/2015 08:52, Honza Horak a écrit :
I'm thinking about repositories and branches structure in CentOS a bit

Currently, there are two ways how components names and branches are
maintained in SCL world:

A) collection name in branch name
we use component name without prefix in dist-git and branches for
specific collections internally in RH, so e.g. mariadb has branches for
mariadb55-*, rh-mariadb100-*, etc..

B) collection name in dist-git repository name
For RHSCL RPMs that are automatically imported into
https://git.centos.org [1], components *use* prefix in repository names
and static set of branches. For example mariadb55-mariadb component
includes branches 'master' and 'c7' [1]. This way seems easier for
dist-git administration and also seems to be preferred way in Fedora [2].

Since we can always add remote branches in (B) and merge/cherry-pick
from those as easy as we can in (A), my question is:

"Are there any practical advantages of (A)?"

I don't see any, but I see some advantages of (B) though, since it seems
to be more consumable to use the same approach (B) in all public repos
and only use (A) internally only for historic reasons.

(A) make sense to me, as the same package/spec can be used in various

Ex: php-pear, various branches
	=> rhel-#
	=> rhscl-##-php54-rhel-#
	=> rhscl-##-php55-rhel-#
	=> rhscl-##-rh-php56-rhel-#

With (B) we need 4 repositories
	=> php-pear
	=> php54-php-pear
	=> php55-php-pear
	=> rh-php56-php-pear

Of course, for historical reason (packages pulled from Fedora where SCL
are not allowed), base packages are really different from SCL packages.
But SCL packages are really similar in the different collections.

But (B) is not really a problem, in fact I mostly never have to
cherry-pick from different branches.

But in some other repo, I use "exactly" the same spec for all targets.

Thanks for the feedback.

One aspect I haven't realized is that with (A) we'd keep information which collections the package is included in. With (B) we wouldn't be able to say that e.g. boost package is used in collections foo42, bar24, ... This seems to be quite useful information.



Does it make sense to you as well?
[no response means yes to me :) ]

[1] https://git.centos.org/tree/rpms!mariadb55-mariadb/c7


SCLorg mailing list
SCLorg redhat com

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]