Re: [Nmh-workers] m_getfld branch merged

[David Levine]

Norm wrote:

> I only understand about every 70% of this discussion, so I
> am talking out of turn, and and maybe nonsense. But to the
> extent I understand the discussion, you guys are investing
> time and effort to make nmh faster.

The bulk of the investment has been to make nmh easier to
enhance going forward.  We need better MIME support and so
on, but needed some (droll, as Paul V. noted) work to the
foundation first.

The attention to performance at this stage is appropriate, I
think.  It will take very little additional effort.  And
given our tools (git bisect), it helps to commit changes
that each do just one thing.  So if such a change causes a
problem, we can quickly isolate it.

And there have been beneficial side effects, including
additions to the test suite and more understanding of
internals.  And resulting deprecation of msh, which will
lead to a bit more cleanup of m_getfld() when it's removed.

>       gone+ has 10702 messages  (9913-24409).

My tests were on folders of about that size and I saw times
of ~20 seconds.  Of course that depends on hardware, but I I
suspect that Linux's caching is affecting your results.  To
clear the caches, before each run:

  # sync; echo 3 >/proc/sys/vm/drop_caches

David