Ken Hornstein <kenh(_at_)pobox(_dot_)com> wrote:
But I have to ask why someone would want to run things this way. BSD
systems - by definition - use flock() in /var/mail. If you run MH with
dot locking for /var/mail, you're eventually going to lose mail, since
the MDA will still use flock(), even if you don't. Not to mention other
tools in ports that will be assuming flock() for /var/mail.
I know the user was on FreeBSD, but a quick googling shows me that there
are a number of places that use dot-locking on the mail spool. Also,
it sounded like Martin was just testing things. Seems reasonable to
make it work.
My MDA (local mailer of sendmail) actually is very nice - it respects
the dot lock put in /var/mail. While I was testing what is going on
having a .lock in the folder nicely put off delivery.
Which actually lead me to another thought: maybe we should
at least _respect_ dot lock (i.e. do not allow writing to a dotlocked
file) but create dot lock only if that particular locking is
requested.
A quick look at the mail.local source code makes me believe
that it actually first creates a .lock file and then proceeds
with flock(). So it uses two kinds of locking at the same time.
//Marcin
_______________________________________________
Nmh-workers mailing list
Nmh-workers(_at_)nongnu(_dot_)org
https://lists.nongnu.org/mailman/listinfo/nmh-workers