Why does checkForbiddenPoint() accept F5 while validating a three?

[HwakYuk]

Could someone clarify why Board::checkForbiddenPoint() accepts this condition while validating a three?

c.pattern(BLACK, dir) == F5

The relevant code is:

if (c.pattern4[BLACK] == B_FLEX4
    || c.pattern(BLACK, dir) == F5
    || (c.pattern4[BLACK] == FORBID
        && c.pattern(BLACK, dir) == F4
        && !checkForbiddenPoint(posi)))
{
    winByThree++;
    goto next_direction;
}

This code is reached only when the original candidate point has:

p == F3 || p == F3S

During lookup-table generation, F3 and F3S are defined through moves that produce F4:

if (patCnt[F4] >= 2)
    p = F3S;
else if (patCnt[F4])
    p = F3;

Therefore, the continuation of an F3/F3S direction should be a move that creates F4, not a move that directly creates F5.

There also seems to be a geometric contradiction.

After the original move X is temporarily placed, the function scans from X through contiguous black stones and tests the first empty point C:

X O O ... O C

Suppose playing C produces F5 in this direction.

There are two possibilities.

1. The exact five includes X

For example:

X O O O C

Playing C gives:

X O O O O

But before C is played, the move at X has already created a four with C as a five-completing point:

X O O O .

That means the original directional pattern at X should have been at least B4, or F4 if another completion point exists. It should not have been F3 or F3S.

2. The exact five does not include X

Suppose C completes five using stones on the side away from X:

X O ... O C O O O O

Since every point between X and C is already black, playing C connects the four stones on the far side to X and the intervening black stones.

The resulting contiguous line contains at least:

X + C + four stones = six stones

and possibly more.

That should be classified as OL, not F5.

So in either case, c.pattern(BLACK, dir) == F5 appears incompatible with the outer requirement that the original direction at X was F3 or F3S.

Is this branch reachable under the current pattern definitions?

If it is reachable, could someone provide a concrete board position?

[dhbloo]

I think your reasoning is correct. This branch is unreachable under the current pattern definitions. F3 and F3S are constructed exclusively from continuations classified as F4, so a continuation in the same direction cannot be F5.
The F5 check is a leftover from an earlier implementation, where this loop also handled double-four validation. After double-four detection was moved to the earlier direct check, this condition became redundant but was not removed.