On 03/12/10 01:15 AM, Alan Coopersmith wrote:
> John Martin wrote:
>
>> On 03/11/10 08:56 PM, Alan Coopersmith wrote:
>>
>>> I don't know why the network& name-service dependencies were in there
>>> previously - John, do you remember?
>>>
>>>
>> That's four years ago, but I recall my concern was lingering
>> logindevperm ownership of the framebuffer device after a reboot
>> where the UID of the owner existed only in the name service and
>> not the local files.
>>
>> However, I'm a bit hard pressed to defend this now.
>>
> Since the service runs as root, does it matter who owns /dev/fb?
> ogl-select just opens it (indirectly via constype), doesn't try
> to check the owner, right?
>
I believe so. This is why I said last night I was hard pressed
to defend it now. Even sleeping on it didn't provide any enlightenment.
I took the two dependencies out of my manifest and could not reproduce
any errors, so I believe they are safe to remove.
> (It might make sense to have something in the startup process run
> the logindevperm functions to chown those back to root just in
> case the machine crashed while they were chowned to a user, but
> that's probably more of something devfsadm should be doing than us.)
>
>
