Re: [Nmh-workers] scan %{body} bug?

[Ken Hornstein]

>I agree (with the re-write, I've never witnessed the problem).

I assume you mean WITHOUT the rewrite, because nobody's re-written
m_getfld last time I looked :-)

It's sort of Steve's setup that triggers it; not only does he have a bunch
of Received: headers, but a huge spam score report, a couple of
Return-Path: headers and a few other bits of extra stuff.  That's enough
with a wider display to put the start of the body right on the stdio buffer
boundary in some cases.  If you don't have that then you'd never see it.

>The code after Ken's excerpt goes directly into the io buffer.  It's in
>the caller, not m_getfld().

Right, I don't think I explained that part well.  While m_getfld() is sort
of under-specified, the basic idea is that you're supposed to keep calling
it until you get a state transition.  From what I can tell everybody does
that EXCEPT scan in this particular instance (the code after that
snipped is used by inc).  So everybody else can deal when m_getfld() returns
a "short" buffer.

Looking at things ... it may be a simple fix, actually.  I wasn't envisioning
any changes to m_getfld(), that's for sure.

--Ken