Location of SetupDaemon in nnrpd.c

Russ Allbery rra at stanford.edu
Thu Jun 13 19:17:29 UTC 2002


Alex Kiernan <alexk at demon.net> writes:

> Ah, but we now use pread() everywhere in cnfs (writes don't, they use
> seek/writev, but thats not a problem), so I think at least from the cnfs
> perspective we could move it back (assuming we're not on a platform
> which uses the pread() emulation in lib/pread.c).

And assuming that pread() doesn't just do a seek and a read inside libc.
Plus, last time I checked, there were major platforms that didn't have
pread().

> That said everything else which is initialised in SetupDaemon would need
> checking (history, overview etc.), plus the other storage mechanisms
> (timecaf looks like it would be the only one which broke from a quick
> inspection).

> Maybe the right approach is to split initialisation for all of these
> subsystems into two parts, one of which is daemon safe, the other not.

Yup, that sounds to me like a good thing to do.  The whole API really
needs an overhaul at some point, though, so timing's always a question.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.


More information about the inn-workers mailing list