question for old rasterization code of stb_truetype.h

[lch32111]

I am studying the old rasterization code of stb_truetype.h by copying the code.
Because it's not used any longer, It doesn't need to be improved. But I want to know my point is correct.

While I was experimenting with it, I found that there was a problematic point:

stb/stb_truetype.h

Lines 2983 to 3006 in f4a71b1

	while (e->y0 <= scan_y) {
	if (e->y1 > scan_y) {
	stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y, userdata);
	if (z != NULL) {
	// find insertion point
	if (active == NULL)
	active = z;
	else if (z->x < active->x) {
	// insert at front
	z->next = active;
	active = z;
	} else {
	// find thing to insert AFTER
	stbtt__active_edge *p = active;
	while (p->next && p->next->x < z->x)
	p = p->next;
	// at this point, p->next->x is NOT < z->x
	z->next = p->next;
	p->next = z;
	}
	}
	}
	++e;
	}

I guess the Sean's intention was not inserting an edge to the linked list of active edges when it reaches the end of stbtt__edge array by putting the sentinel's y coordinate like this:

stb/stb_truetype.h

Lines 2939 to 2940 in f4a71b1

	e[n].y0 = (off_y + result->h) * (float) vsubsample + 1;

However, the scan_y variable is updated like this:

stb/stb_truetype.h

Lines 2943 to 2945 in f4a71b1

	for (s=0; s < vsubsample; ++s) {
	// find center of pixel for this scanline
	float scan_y = y + 0.5f;

So, the y coordinate of the sentinel coudl be less than scan_y if the variable vsubsample is large enough.
So I guess it will be safer to directly check whether it reaches the end of the stbtt_edge array like this:

// init code
stbtt_edge* sentinel = e + n;
...
while (e != sentinel &&  e->y0 <= scan_y)