Re: [RFC/PATCH] Fix "grep -w"

[Junio C Hamano <junkio@cox.net>]

"Morten Welinder" <mwelinder@gmail.com> writes:

> 1. Are you sure that going to the end of the first match is correct?
> It seems to me that this will skip matches.  Say you search
> for ".*" on a line that reads
>   " xxx".

It is fine for your example, I think.  .* matches the entire
line the first time, and BOL and EOL are defined to be word
boundaries.  But you are right.  If the pattern is "x xx* x" and
the line is "x x xx x", the first round would match the first 5
bytes, we find that 6th byte 'x' makes it not a word boundary,
and redoing the match starting at 6th is a wrong thing to do.
We should find "x xx x" starting at the 3rd byte.

> 2. What about "^"?

The pattern would not match the second time anyway, so I do not
think it is such a big deal.

But there is another bug I just spotted.  git grep -w -e '^x'
matches line "xxx" (when not cheating with external grep).

> 3. What about empty matches?  That could take a while...

True.  So we would need to make sure we advance at least one.