| 1 |
On Wednesday 20 July 2011 14:16:06 Grant Edwards did opine thusly: |
| 2 |
> On 2011-07-20, Mick <michaelkintzios@×××××.com> wrote: |
| 3 |
> > Every time after you emerge xorg you're meant to remerge its |
| 4 |
> > drivers (evdev being one of them). Usually there is some elog |
| 5 |
> > message telling you to run |
| 6 |
> > |
| 7 |
> > qfile to find what is need to be reinstalled: |
| 8 |
> > qlist -I -C x11-drivers/ |
| 9 |
> |
| 10 |
> I've always wondered why, if portage knows that has to be done, |
| 11 |
> can't portage just go ahead and do it? |
| 12 |
|
| 13 |
Actually not. Portage doesn't know it has to be done, the ebuild dev |
| 14 |
knows and stuck a literal message in one of the post_() functions. |
| 15 |
|
| 16 |
To do what you mention, the driver would have to depend on xorg, but |
| 17 |
in fact xorg-server depends on the xorg-drivers meta package, which in |
| 18 |
turn depends on the drivers defined in USE. |
| 19 |
|
| 20 |
In theory, one could use PDEPEND I suppose, but I also suspect that |
| 21 |
would cause circular dependencies. The other option is to put an |
| 22 |
emerge call in pkg_postinst() but that would be heavily frowned upon |
| 23 |
as |
| 24 |
|
| 25 |
- it's a highly unusual solution and just an ugly hack |
| 26 |
- it wreaks the dep graph that emerge generates |
| 27 |
- stuff will get emerged that isn't visible at the beginning |
| 28 |
- portage can't definitively tell you before you start that the |
| 29 |
drivers will be remerged |
| 30 |
|
| 31 |
|
| 32 |
-- |
| 33 |
alan dot mckinnon at gmail dot com |
Archives