Breakage on OS X

[Harald Hanche-Olsen]

I just built emacs from the latest commit on the git mirror at
savannah – this one:

commit 5c1915d10b3716879785fe49f5cfe20beeb37090
Author: Paul Eggert <address@hidden>
Date:   Thu Apr 17 07:59:22 2014 -0700

and I am seeing some really odd behaviour.

I'd file a bug report, but I have been unable to find a reasonably
reliable way to reproduce the problem. It just happens at random
times, just often enough to be really annoying.

The most common symptom is that C-e moves point to the end of the last
line of the buffer (before the final newline) instead of the current
line. Once this happens, it keeps happening in that buffer, but only
if the current position is past some random (but fixed) point in the
buffer. When this happens, C-k will kill to the end of the buffer, and
(forward-line 1) will move point to the end of the buffer.

Now, forward-line seems pretty basic: It is written in C, and just
counts newlines. So it appears that something is broken at a fairly
fundamental level. Just to be sure, I have checked that all the
newlines that are skipped over really are still proper newlines, and
that (eolp) returns t when expected.

When the buffer enters this state, other symptoms often appear too,
such as wrong indentation; either way too much, i.e., 30 to 40 columns
too much, or no indentation at all.

I have seen this happen most in HTML mode buffers, but as I was going
to start this email, it happened in the message buffer as well.

I must add that I recently upgraded xcode from the app store, which
makes me suspect a compiler bug rather than an emacs bug.

It would be interesting to know if anybody with the latest OS X
(10.9.2) and the latest xcode (5.1.1) sees the same problem? Or if the
latest commit runs without problem when compiled with the previous
xcode. (Unfortunately, downgrading xcode to find out is really
difficult.)

– Harald