Comment 23 for bug 289216

After some "quality time" tinkering with pointer sizes and xev, I found a fairly solid pattern to the areas which trigger flickering:

1) Display size is 2560x1024, DPI is 96.

2) Flickering appears to occur when *any* portion of the pointer image is withing the affected area. The effect is distinctly noticeable on a large dark arrow, since part but not all of the cursor can end up with flickering, if positioned correctly.

3a) On the *primary* screen (LVDS on the T60) the areas are roughly 36 pixels high, and their upper bounds appear to follow the pattern (N*79)+40. I initially guessed at (N*80)+40, but when I checked near the bottom of the screen, it was off by a good ten pixels or so, while 79 appears to predict every value my screen can reach. Also, 1024 is an integer multiple of the formula ((12*79)+40+36 = 1024).

3b) On the *secondary* screen, the flickering appears to be a single column from the top of the screen to the bottom; there are no obvious "breaks" of space where flickering does not occur.

4) On both screens the flicker area appears to be roughly 8 pixels wide, and the right-most bound appears to align fairly closely with (N*512). Much like (3a), this means that the right-most part of the final area aligns pretty much exactly with the right-hand side of the screen (5*512=2560).

5) As a final note, in a couple of spots the visual result of the flickering was to have multiple 'ghost' copies of the pointer offset directly below the main pointer. This was not consistent between areas, though any given area appears to produce roughly the same pattern each time it flickers, even when the pointer has moved away in the interim.