Re: [gentoo-dev] [gentoo-portage-dev] About forcing rebuilds of other packages issue

[Pacho Ramos]

El mié, 06-06-2012 a las 02:51 +0200, Michael Weber escribió:
[...]
>
> [1] if you forget the -X on module-rebuild, you might no longer have
> the virtualbox-modules version installed in the tree (no packages
> satisfy ...). virtualbox does remove old versions real quick.
>
> The fun part comes with non-root users trying to log in:

Yeah, I also had a similar problem with nvidia-drivers, maybe
module-rebuild should default to "-X" behavior, or is there any reason
why forcing the current behavior is better? Do we really should support
by default setups that don't apply all updates (neither locally mask
unwanted newer versions) after syncing their tree?

>
> [2] You've updated nvidia-drivers (kernel module providers in general)
> userland and kernel modules, but forget to `rmmod nvidia`, or you
> can't without terminating user sessions, it impossible to start new X
> servers due to version mismatch between userland and kernel (applies
> for virtualbox as well)
>

Maybe if we were able to call "rmmod -w nvidia" from nvidia-drivers
ebuild... that way, once you log out from X, old module would be
outloaded and new one loaded by X when restarting. The problem is that
there is no way to run this command after emerge "automatically"

> [3] You've updated zlib, but failed to recognize it in the emerge -av
> output. You get angry reports about broken luatex and inkscape
> (imagemagik) because of some nasty zlib abi version mismatch, hidden
> from revdep-rebuild.
>
> [5] lafilefixer (funny)

I am not sure if this is still needed these days :-/, at least portage
looks to fix them, but I think this is not supported on other PMs (or
maybe they handle this other way apart from lafilefixer also) 

> [4] python-updater (rare)
> [6] ocaml gets broken after update w/o lablgl rebuild
>   https://bugs.gentoo.org/385869
>
> Well, I'm lazy, and do this in the backgound, half asleep.
> And I admit that [1] and [2] are my faults, but [3] is very annoying
> (just like libdl related stuff) and esp. kernel+module updates take a
> lot more than just a few 'REBUILD' packages.
>
> Is there any chance to detect this ZLIB_VERSION problem with
> revdep-rebuild (worst case: add a list of possibly broken packages
> with tests)?
>
> =====
>
> I understand the urge for `eupdate` but that needs an agreement on
> the implementation, and I see some rought edges here, where unattended
> script magic most likely fails.
>
> Michael -- half asleep
>
> - --
> Gentoo Dev
> http://xmw.de/