bug#16433: Test case for newline cache corruption

[Eli Zaretskii]

> From: Ken Olum <kdo@cosmos.phy.tufts.edu>
> Cc: 16433@debbugs.gnu.org
> Date: Tue, 27 May 2014 12:20:52 -0400
> 
> Here is some additional output which perhaps will be helpful.

Thanks.  I know why I couldn't reproduce what you see: you've sent me
the wrong test.mbox file.  The one you are using has only 1 message in
it, whereas the one you sent me had 2 messages, probably because you
have already pressed F4 once to read a new message.

If I manually remove the 2nd message from test.mbox, I see the same
behavior as you reported.  Well, almost the same: for some reason the
traces from region-cache.c are very similar, but not exactly
identical.

I deduced that the initial mbox was not the one you used from the
traces you sent: they clearly indicated that your initial mbox was
smaller than mine (the last number in know_region_cache trace is the
buffer's point-max).

In any case, I seem to have now a reproducible recipe, and can work on
it locally.  Thanks so much for your help.