Re: [Nmh-workers] new command lacks lock

[Ralph Corderoy]

Hi Robert,

> > What if the `rm 2' was temporary with the script about to re-instate
> > it after deleting 4 from the sequence?
> 
> And what if some completely different file was moved in and called '2'
> before the mark -d ?   That one would remain in the sequence, though
> it has no business being there really ...

I think it does since I'd have moved it in with an understanding of how
sequences aren't automatically updated as if each message contained a
sequence membership list.

> I think the moral is that if you go about playing with the MH
> datastore without involving MH commands, then it is your
> responsibility to make sure everything remains consistent

I disagree.  mark(1) doesn't mention scanning the directory to re-build
a sequence's members, especially not on the delete of a single item from
a sequence.  I did arrange for everything to be consistent as far as I
knew how mark worked from the documentation and a possible model of its
implementation.  :-)  How it actually works isn't great on directories
with many entries, especially if one wants to, e.g. move messages
between sequences;  multiple runs of mark.

Cheers, Ralph.