| 1 |
On Saturday 04 April 2009 08:59:22 Thomas Sachau wrote: |
| 2 |
> i would like to hear about other opinions about real multilib support |
| 3 |
> within our tree and package managers. From what i know, there are mainly 2 |
| 4 |
> different ideas: |
| 5 |
> |
| 6 |
> 1. Do the main stuff in the package manager (e.g. if the ARCH is amd64 and |
| 7 |
> the package has x86 keyword, the package manager adds a lib32 useflag, |
| 8 |
> which would additionally install the 32bit variant of that package together |
| 9 |
> with the normal 64bit install). |
| 10 |
> |
| 11 |
> pro: -much lesser work for package maintainers |
| 12 |
> |
| 13 |
> contra: -needs addition in PMS and support in the pms, which will need some |
| 14 |
> work on their side |
| 15 |
|
| 16 |
get a *working* implementation first and *then* worry about specing it. once |
| 17 |
you have something running with portage, the spec should fall naturally out. |
| 18 |
previous multilib methods attempted to spec things out without any real code |
| 19 |
and they've all just died. |
| 20 |
|
| 21 |
> 2. Do the main stuff in the ebuilds themselves (e.g. an additional eclass |
| 22 |
> multilib-native.eclass, any ebuild with 32bit support would then need |
| 23 |
> adaption and of course inheriting that eclass) |
| 24 |
|
| 25 |
this is dead end and useless overhead, and i would reject it from any core |
| 26 |
package someone would try to merge. |
| 27 |
-mike |
Archives