Tim Goodwin writes:

> Thanks for the report.  I can reproduce this.
> 
> It's not been reported before, so I don't know what's going on, but I'll
> let you know as soon as I have a fix.

I did send something to to mailing list on Saturday, but it hasn't
gotten through yet. I seem to have figured out the use of m in a List
structure: if there are no metacharacters, NULL, and if there are, we
put a \0 where there's a non-metacharacter in the corresponding part
of n, and a \001 where there is.

Now in lmatch() when we go through looking for '*'s, we have this
loop:

	for (i = 0; p->w[i] != '\0'; i++)
		if (p->w[i] != '*' || p->m[i] != 1) {

Which assumes that m is as big as w (i.e, not NULL). I came up with
this replacement, which works for all the test cases I can think of:

        if (s == NULL) {
                if (p == NULL) /* null matches null */
                        return TRUE; 
                for (; p != NULL; p = p->n) /* one or more stars match null */
                        if (p->w && strspn(p->w, "*") == strlen(p->w) &&
                            p->m && strspn(p->m, "\001") == strlen(p->m))
                                return TRUE;
                return FALSE;
        }

I was also able to get rid of 'okay', which i found hard to follow.
The comment about the null string is gone too, because IMO the null
string fails for the same reason 'foo' fails. There's really no need
for an explicit check.

A patch is attached. Out of curiosity, why was the decision made to
use '\000' and '\001' instead of, say, '0' and '1'? For example in the
old code, we have "p->m[i] != 1", and it wouldn't be that hard to
write "p->m[i] != '1'" instead. I'm sure there's a reason, but I can't
guess it.

-- 
Written with 100% free software. Please support the following websites:
www.debian.org www.noamazon.com www.gnu.org www.opendvd.org lpf.ai.mit.edu