VMiklos wrote:
On Wed, Mar 08, 2006 at 11:40:26PM +0100, Aurelien Foret <aurelien@archlinux.org> wrote:
Finally, I think pacman 3.0 is the right moment to implement it.
hm, not sure about this. the libarchive patch was rejected as "we have enough changes for 3.0". i think this backend change is much more problematic than the libarchive one...
You can hardly compare the libarchive and the backend things. Replacing libtar by libarchive is purely an internal change. Users won't even notice it. It could be done at any time, in any release. It does not matter. I agree to say that the backend is more problematic: it will have impacts for developers wanting to build something based on the library. It has to be designed correctly before releasing it. Any change in its interface within alpm will mean upgrades and issues for developers implementing their own backends.
what about first just do the TODO, release 3.0, and _after_ start to work on this?
I don't think it is respectful for pacman users and more especially for developers to release 3.0 which will somehow break the sync db compatibility (although it should be hardly noticeable), and then to release a new version with a new breakage. Let's break it only once, especially since the modification I'm suggesting does not produce more annoyance than a release without it. The shift from 2.x to 3.0 is a good time for such things. Anyway, my first mail was maybe not clear, but there's only one thing my patch is about: implementing correctly the sync database upgrade, by taking the opportunity of pacman 3 release to do it without additional troubles for users. This will get ride of the "ORE" tag in alpm.c:alpm_db_update. And AFAIK, _this_ is on the TODO. -- Aurelien