From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 03 Aug 2011 17:30:27 GMT
Subject: Re: [PATCH 17/18] revert: Introduce --continue to continue the operation
Message-ID: <7v4o1ycsi4.fsf@alter.siamese.dyndns.org>
In-Reply-To: <CALkWK0mrx+3jdCQD9xya3AWMsPpSiZzEz+Z9XVxZNzw3UdKMVw@mail.gmail.com>

Ramkumar Ramachandra <artagnon@gmail.com> writes:

>>> +     for (p = buf.buf; *p; p = strchr(p, '\n') + 1) {
>>
>> This relies on a "\n" at the end of the last line...
>
> Yes, that was intentional.  Every editor I know of inserts a '\n' at
> the end of the last line,...

Why do you think "diff" has a special codepath to report "No newline at
end of file"? Because some editors do leave an incomplete line incomplete.

When working on a "buf" with an incomplete line at the end, the last round
of the loop will assign "(char *) 1" to "p", try to see "*p" is NUL and
segfault, no?