I don't think CScrollView::DoMouseWheel can handle scroll sizes less than one line. I think that was also an issue with the scroll messages created by touch gestures, in which case explicitly implementing touch support fixed the issue but WM_MOUSEWHEEL message with smaller scroll amounts will remain problematic.

Thanks. I'll look into it. I may be able to come up with something.

I have a working fix. I'll create a pull request on GitHub.

PR created: https://github.com/OpenMPT/openmpt/pull/15

The code looks sensible, but I suppose this also affects horizontal scrolling (e.g. when there's lots of pattern channels)? Or is the different implementation of CModScrollView::OnMouseHWheel causing that to work? If so, taking the horizontal scrolling code and reusing it for vertical scrolling would be a one-liner fix.

I've never seen a mouse device with fine-grained horizontal mouse wheel (just tilt left/right, which are typically binary switches). However, the MFC base class automatically converts vertical mouse wheel rotation into horizontal scrolling when there's no vertical scrollbar, so fine-grained horizontal scrolling via vertical mouse wheel rotation is possible. And yes, it is still broken even with my fix.

CScrollView::DoMouseWheel contains the defect/limitation, but CModScrollView::OnMouseHWheel doesn't use it, so a somewhat different form of my fix is needed here -- especially because horizontal scrolling is only permitted to occur in whole-column increments.

I'll get to work on a revised fix.

Pull request updated. Please review.

I tested the updated code now, and it breaks the intended horizontal scroll behaviour (direction is inverted and one mouse wheel notch jumps by two channels instead of one - probably because the truncation now happens in a different place).
I think the attached patch is simpler and may also solve your problem. Can you give it a try?

Index: Globals.cpp
===================================================================
--- Globals.cpp	(revision 18395)
+++ Globals.cpp	(working copy)
@@ -647,23 +647,32 @@
 }
 
 
+static short RoundMouseWheelToWholeStep(int value)
+{
+	if(value > 0)
+		return mpt::saturate_cast<short>(mpt::align_up(value, WHEEL_DELTA));
+	else
+		return mpt::saturate_cast<short>(-mpt::align_up(-value, WHEEL_DELTA));
+}
+
+
 BOOL CModScrollView::OnMouseWheel(UINT fFlags, short zDelta, CPoint point)
 {
 	// we don't handle anything but scrolling just now
-	if (fFlags & (MK_SHIFT | MK_CONTROL)) return FALSE;
+	if(fFlags & (MK_SHIFT | MK_CONTROL))
+		return FALSE;
 
-	//if the parent is a splitter, it will handle the message
-	//if (GetParentSplitter(this, TRUE)) return FALSE;
-
 	// we can't get out of it--perform the scroll ourselves
-	return DoMouseWheel(fFlags, zDelta, point);
+	return DoMouseWheel(fFlags, RoundMouseWheelToWholeStep(zDelta), point);
 }
 
 
 void CModScrollView::OnMouseHWheel(UINT fFlags, short zDelta, CPoint point)
 {
+	zDelta = RoundMouseWheelToWholeStep(zDelta);
+
 	// we don't handle anything but scrolling just now
-	if (fFlags & (MK_SHIFT | MK_CONTROL))
+	if(fFlags & (MK_SHIFT | MK_CONTROL))
 	{
 		CScrollView::OnMouseHWheel(fFlags, zDelta, point);
 		return;

Sorry for the delayed response. Between holiday travels, a 2 week bout of COVID, layoffs and reorgs at my employer, and my personal PCs all getting jacked up by upgrading them to Windows 11, I've been pretty busy dealing with other things.

Yes, I'll give your patch a try soon and report back.

Did you get a chance to try my patch yet?

Sorry, not yet. I've had a lot going on lately. It's still on my radar though.

I had to replace my laptop and now I have access to a precision touchpad. I can see that my patch fixes the issue but now scrolling is a bit too fast / sensitive. I'll investigate a bit more.

r19442 should work better, can you check?