unintended consequences from (I believe) 4.76 change #10

Change 10 in MUSHclient 4.76 says


This was done because of http://www.gammon.com.au/forum/?id=11197

Except now there are far worse problems because it's not just the mouseover callback firing, but rather an entire mouse-move event happening any time a hotspot gets created. It looks like every time a trigger fires in any plugin the mouse position gets reset, which manifests in the following way:

I have a simple resizable miniwindow, like in this picture https://aardwolfclientpackage.googlecode.com/svn/trunk/wiki_images/ascii_map1.png

/While/ I am dragging the resizing tab, if any window updates its contents it does terrible things.

First, if as a result of the mouse-move the current miniwindow recreates hotspots (which then calls mouse-move again) then this obviously causes an infinite loop in the window (this happens in my chat capture window plugin, for example). But it also appears that if any /other/ miniwindow creates hotspots then the stored mouse position can get messed up which makes the window that is currently being resized jump around each time.

I originally was looking for /only/ mouseover callbacks to fire, which I think would solve these problems.

Yes, but...

The mouseover callbacks are supposed to be paired, that is you get a mouseover followed by a cancel_mouseover. The function that I called (which happens on a mouse move) should have kept the original integrity, that is it cancels any previous mouseover, and starts a new one, etc.

Let's say for example you created a hotspot in response to a mouse-down, then the mouseover should not be generated because we are currently mousing-down.



Well you wanted a mouse-move generated when you created a hotspot so I supposed you might expect that.



I don't see how that fixes the problem of the loop.

Would it help if there was an extra test? That the extra code is only executed if the mouse is (thought to be) over the newly created hotspot (rather than unconditionally)?

I think it is a bit of a design flaw, to make it so that if the mouse moves, you create a hotspot, the hotspot being one that detects mouse moves and recreates itself.

Would it help if I made it impossible to create hotspots in a mouse-move event?

One problem is the issue of "which mouse position?".

If the mouse is moving fast we have at least 3 positions:

1. Where the mouse was when the mouse move event was generated by Windows.
   
2. Where the mouse is when the mouse-move plugin callback is called.
   
3. Where the mouse is when you call WindowAddHotspot.

They could all be different positions.

I don't want a mouse-move generated when a hotspot is created. I don't actually think that generating a mouse-move is correct or appropriate, because the mouse may not in fact be moving. It is the hotspot that is moving rather than the mouse, which is why I think that /only/ onmouseover should be called (if appropriate).

I don't know, maybe, but see above.

My understanding of the current problem is that this is not actually what is happening. In fact, I don't think that that's even possible, since hotspot creation tends to delete callbacks. It looks like the mouse-move event is happening even if an entirely different miniwindow creates hotspots.

All I know is that with 4.75 all of my resizing works great. And in 4.76 if any miniwindow other than the one I'm trying to resize updates its hotspots because of a trigger firing the resizing gets screwed up.

I seem to have done a couple of things wrong there.

The code for generating the mouse-move was using the wrong mouse coordinates which would tend to explain what was happening to you. That is fixed in version 4.77

Also code to detect infinite loops was added. Dropping that on top of your Aardwolf build it seems to work OK now (without any code changes on your part).