nmh-workers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Nmh-workers] MH-W intro/help request


From: Ken Hornstein
Subject: Re: [Nmh-workers] MH-W intro/help request
Date: Thu, 04 Dec 2014 11:05:21 -0500

>If I understand your question:  sequence names must start with a
>letter.  Fortunately.  That's documented in mh-sequence(5) and
>enforced by the code:
>[...]

I was just wondering, since the sequence checking code was first in
m_convert().  I just checked; if you create a numeric sequence by hand,
yes, you totally can use it!

>pick now calls folder_read() twice:  could we remove one of
>them, at least sometimes?  The first read is without
>locking, which is good.  But I don't know if we can reliably
>bypass the second one for cases where the sequences file was
>already up to date.  And I don't know if it'd be worth the
>complexity to try.

That was done because as part of the locking rework because either a) we
have to hold a sequence file lock during the whole run of pick, or b) we
have to reread the sequence file after pick is done because the sequence
file might have changed during the run of pick (we do something similar
in inc).  I'm assuming we still want to keep sequence files consistent.

But ... it occurs to me, looking at things now, we really only need to
do either of those things if we're using pick to add to the sequence file.
pick does not set the Previous-Sequence, so normally the sequence file
does not need updating.  So the default case is you don't need the
second call to folder_read().  I think what we should do here is:

- If you don't give the -seq flag to pick, don't make the second call to
  folder_read().

- Create a new option to control whether or not you want to maintain
  a lock during the whole run of inc and pick.  I say default to NOT
  having the lock during the full run of these commands and have two
  calls to folder_read(); people with 100K+ messages in a single folder
  could enable this flag.  I do not know what to call this flag; -fulllock?
  Too many 'l's in a row, for one.

What do others think?

--Ken



reply via email to

[Prev in Thread] Current Thread [Next in Thread]