Re: [PATCH] diff: avoid stack-buffer-read-overrun for very long name

From: Bert Wesarg <bert.wesarg@googlemail.com>
Date: Apr 26, 2012, 16:21 UTC
Message-ID: <CAKPyHN1mFGiZd7dDH-stUmr3H1JHwxcP1DkjCYNXZd6Bt-P7+w@mail.gmail.com>
In-Reply-To: <xmqq62cma2uo.fsf@junio.mtv.corp.google.com>

On Thu, Apr 26, 2012 at 18:13, Junio C Hamano <gitster@pobox.com> wrote:

Show 19 quoted lines

> Jim Meyering <jim@meyering.net> writes:
>
>> What do you think about replacing those two append-if-needed two-liners:
>>
>>     if (buffer2.len && buffer2.buf[buffer2.len - 1] != '/')
>>             strbuf_addch(&buffer2, '/');
>>
>> by something that readably encapsulates the idiom:
>>
>>     strbuf_append_if_absent (&buffer2, '/');
>>
>> (though the name isn't particularly apt, because you might
>> take "absent" to mean "not anywhere in the string," so maybe
>>   strbuf_append_if_not_already_at_end (ugly) or
>>   strbuf_append_uniq
>> )
>
> I am not good at names, but strbuf_terminate_with(&buffer2, '/')
> perhaps?

strbuf_ensure_terminator(struct strbuf* buf, int term, int always)?

Show 14 quoted lines

>
>> There are several other uses that would benefit from such a transformation:
>> To find the easy ones, I ran this:
>>
>>   git grep -B1 "strbuf_addch.*'"|grep -A1 '!='
>>
>> I've manually marked/separated the ones that don't apply.
>>
>> ------------------------------------
>> builtin/branch.c-     if (!buf.len || buf.buf[buf.len-1] != '\n')
>> builtin/branch.c:             strbuf_addch(&buf, '\n');
>> --
>> strbuf.h-     if (sb->len && sb->buf[sb->len - 1] != '\n')
>> strbuf.h:             strbuf_addch(sb, '\n');

Please note, that while they are checking the .len, they both behave differently if .len == 0 or not. The first always append a '\n', the latter only, if the string isn't empty.

Bert