| 1 |
Ulrich Mueller <ulm@g.o> wrote: |
| 2 |
> |
| 3 |
> Other problems appear, though. Documentation is installed in a ${PF} |
| 4 |
> subdir, so install locations actually do change when updating the |
| 5 |
> minor revision. |
| 6 |
|
| 7 |
Yes, the minor revisions should not be exported into the variables |
| 8 |
of ebuild.sh. I had forgotten this. |
| 9 |
|
| 10 |
> Also some method for updating binary packages would be needed. |
| 11 |
|
| 12 |
I already mentioned that. But even if this is not implemented, |
| 13 |
this is only a minor issue. |
| 14 |
|
| 15 |
> All in all, I'm not convinced if the cure wouldn't be worse than the |
| 16 |
> disease here. |
| 17 |
|
| 18 |
The disease is making the distribution almost useless |
| 19 |
or having broken dependencies. |
| 20 |
|
| 21 |
> It would introduce another level of complexity, in order |
| 22 |
> to avoid a few rebuilds. |
| 23 |
|
| 24 |
It seems you never counted the number of silent modifications |
| 25 |
to the tree: Just compare the number of changed packages in |
| 26 |
metadata/ with the number of packages shown by eix-sync... |
| 27 |
I would guess it means roughly that you have to recompile your |
| 28 |
whole installation once a week, 95% of the rebuilds being due to |
| 29 |
not fixing the package manager to work properly with dynamic deps. |
| 30 |
|
| 31 |
Actually, I still think that implementing dynamic deps correctly |
| 32 |
would be better, but minor revisions do not exclude this |
| 33 |
and are probably simpler to implement. |
Archives