[PATCH] remove usage of g_assert() in blockdev

Mark Rosenstand mark at borkware.net
Wed Nov 15 00:27:33 PST 2006


On Wed, 2006-11-15 at 00:15 -0500, David Zeuthen wrote:
> On Tue, 2006-11-14 at 15:43 -0600, Steev Klimaszewski wrote:
> > Not to mention the dependency of HAL on PolicyKit, which requires pam... 
> > or am I missing something here?
> 
> (PAM already lives in /lib, at least on Fedora if not everywhere else.)
> 
> PolicyKit is used as a D-Bus service and is only used once callers start
> calling into HAL. 
> 
> At a future point, when D-Bus system bus activation lands, PolicyKit
> will probably just be started only if there is a need for it, e.g. when
> HAL or others call into it... And then the daemon can exit too e.g. when
> it haven't provided service to anyone for e.g. 1 or 2 minutes.
> 
> Either way, having / and /usr on separate file systems doesn't really
> make sense in this millennium I think. As such, moving HAL out of /usr
> does not make any sense to me. I think we simply just want to advise
> distributors that / and /usr needs to be on the same file system. 

Then why even bother with /usr? (--exec-prefix=)

Anyway, the current FHS revision is pretty clear regarding this topic
and I think hal would be a bad citizen to ignore that and even recommend
distributors to do the same. Proposing such changes for future revisions
would probably not be as efficient (the subject would have to be
discussed and argumented) but in the end, it makes a better OS for all
of us when the different parts of the system agree on these things.

Then again, we know how the xorg git move was decided and it looks like
that went OK. :)

> (Of course there will always be distributions and technology enthusiasts
> who insist on / and /usr being separate file systems - the answer here
> will be that they need to run the udev coldplug process twice. Once
> right after the initramfs -> real user space transition, and once as
> part of HAL startup. That is _essentially_ what happens today anyway.)

As long as that's possible (i.e. the coldplug code isn't removed from
udev), at least I'm happy :)



More information about the hal mailing list