Re: [hatari-devel] One bug, one warning, one :)

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

On 14/11/2012 00:17, David Savinkoff wrote:
> Attachment has comments.


Hello,

although you got different results after recompiling, here's what I see 
from Hextracker in my case (no mouse problem for me) :

At start, it sends :

IKBD_Cmd_TurnMouseOff
IKBD_Cmd_DisableJoysticks
IKBD_Cmd_RelMouseMode

Then, you get (0xf8 dx dy) sequences while moving the mouse.

Once "load mod" is pressed, this gives :

send byte=0xfa delay=0		<- LMB pressed
send byte=0x00 delay=0
send byte=0x00 delay=0

Hextracker "restores" the GEM by putting back the ikbd in its usual GEM 
mode :

IKBD_Cmd_TurnMouseOff
IKBD_Cmd_ReturnJoystickAuto
IKBD_Cmd_RelMouseMode

Then file selector is opened and I release LMB

send byte=0xf8 delay=0		<- LMB released
send byte=0x00 delay=0
send byte=0x00 delay=0


In your traces, we have :

send byte=0xfa delay=0		<- LMB pressed
send byte=0x00 delay=0
send byte=0x00 delay=0

Then those bytes look like bytes sent from IKBD_SelAutoJoysticks (which 
is called by IKBD_Cmd_ReturnJoystickAuto)

send byte=0xff delay=0
send byte=0x01 delay=0
send byte=0xfe delay=0
send byte=0x80 delay=0
send byte=0xfe delay=0
send byte=0x00 delay=0

In my case, I don't have those joystick bytes ; do you have a joystick 
plugged, or did you accidentally press alt+J ?
Those byte should not be sent if there's no joystick.

Maybe there's a memory leak hidden somewhere that was never seen before 
and that is triggered in your case with you own gcc setup, but I don't 
see how to track this (it could also be a gcc bug, but a bug in Hatari 
is more likely).

As Eero said, maybe running with valgrind and/or putting a breakpoint in 
gcc in IKBD_SelAutoJoysticks() would give some hints on what conditions 
allowed those 0xff and 0xfe bytes to be sent ?

Nicolas