https://gitlab.synchro.net/main/sbbs/-/commit/c9737b87032bbed1c4e4d2c6  
  Modified Files:  
   src/doors/termgfx/sgrmouse.c src/doors/termgfx/test/test_mouse.c  
  Log Message:  
  termgfx: the motion bit is authoritative -- 97 is a moving pointer  
    
  A wheel notch carrying the motion bit was rescued as wheel-down, on the  
  reasoning that a hover never sets the low bit and on a session measured at  
  75 such reports against 30 clean ones. A later capture inverts that ratio  
  completely: 1492 clean 64/65 detents against 16 of 97, several of the 97s  
  arriving 6 to 8 ms apart with the coordinates walking one cell at a time.  
  No hand turns a wheel that fast; that is a pointer being moved.  
    
  What the rescue actually bought was a phantom scroll on every pointer  
  movement. A door's block selection changed whenever the mouse moved, and  
  behaved correctly only while the hand was held perfectly still -- which is  
  exactly how it was reported, and was read as a terminal bug for several  
  rounds.  
    
  This corrects 1da9027368 (fire-20-buys, 2026-08-09), which added  
  the rescue. Real detents arrive as 64 and 65 without the motion bit, in  
  their thousands, so trusting the bit loses nothing. The test now pins both  
  directions: 96 and 97 are motion, 64 and 65 are notches. It fails against  
  the rescued decode.  
    
  Co-Authored-By: Claude Opus 5 (1M context)   
  --- SBBSecho 3.37-Linux  
   * Origin: Vertrauen - [vert/cvs/bbs].synchro.net (1:103/705)