solaris /dev/console patch

[Robert McQueen]

Havoc Pennington wrote:
> Well, I think things can be more or less cleanly done, or done
> differently (separate files, runtime selection, single huge function
> with #ifdefs)
> 
> I'm just saying, let's not create a generic plugin architecture or
> something ;-) it's really not that much code in question.

Sure, I just wanted your input on the preferred mechanism before we
accept too many more patches which tend towards the single huge function
with #ifdefs. :)

> Unless you cache the separate process result I think it'll be on the
> slow side... potentially called once per message, no? Not that anything
> using this is necessarily time-critical right now.

Well, the current foreground user is reasonably time-critical, because
another user can take over at any time. The longer we cache this
information for, the closer we end up turning the foreground policy into
just "at console", or worse "was once at the console". If we do cache
it, it should be for a reasonably short period. I'd prefer we
implemented the policy in a correct manner and that the users of it be
aware that by definition, this policy carries a certain overhead. On the
other hand, on the platforms where we're doing somehing like forking a
seperate process (Linux) then forking is pretty quick anyway. :)

> Havoc

Regards,
Rob