| 1 |
On 13:49 Sun 16 Jan , Ciaran McCreesh wrote: |
| 2 |
> People seem to have started using blockers with package moves recently. |
| 3 |
> For example, if cat/a is being moved to cat/b, people have started |
| 4 |
> putting !cat/a as a dependency in cat/b. This is bad, for two reasons. |
| 5 |
> |
| 6 |
> First, you shouldn't have to do that. If package moves aren't working, |
| 7 |
> we've got bigger problems, and throwing in some blockers won't help |
| 8 |
> there. |
| 9 |
> |
| 10 |
> Second, when performing updates, Paludis also rewrites dependencies of |
| 11 |
> installed packages to use the names. This means that your block on |
| 12 |
> cat/a will be rewritten to a block on cat/b, which means the package |
| 13 |
> ends up blocking itself. |
| 14 |
> |
| 15 |
> We've got two options here. |
| 16 |
> |
| 17 |
> Option the first is that people stop writing stupid blockers on package |
| 18 |
> moves. Unless someone can come up with a convincing reason to keep them |
| 19 |
> there, this is the option that should be taken, and repoman should |
| 20 |
> enforce it. |
| 21 |
> |
| 22 |
> Option the second is that I make Paludis stop rewriting blockers for |
| 23 |
> package moves. This is bad, because it means legitimate blockers which |
| 24 |
> should be honoured will end up disappearing. |
| 25 |
> |
| 26 |
> Does anyone care to justify their "block the old name" habits? |
| 27 |
|
| 28 |
How about playing nicely with overlays where the moves didn't happen |
| 29 |
(yet)? |
| 30 |
|
| 31 |
-- |
| 32 |
Thanks, |
| 33 |
Donnie |
| 34 |
|
| 35 |
Donnie Berkholz |
| 36 |
Sr. Developer, Gentoo Linux |
| 37 |
Blog: http://dberkholz.com |
Archives