Re: [hatari-devel] Possible problem with Pexec emulation

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 19.11.2022 7.10, Thorsten Otto wrote:
> Which looks even more strange, since the bus error seem to be from the tas
> instruction, but the move immediately before that was executed successfully?
> And why should a5 suddenly point to $e00000 instead of the blitter address? Or
> is that address maybe because the blitter tries to write $e00000?
Tell Hatari debugger to track address history:
> history cpu 1024

And ask it to invoke debugger on bus errors:
> setopt --debug-except bus

When debugger gets invoked on the bus error, ask it to show last 
executed instructions:
> history 32

You could also enable profiling to get better picture of rest of the 
code running before bus error (and a backtrace, if Smurf binary includes 
debug symbols):
> profile on


Profiling will show how many each instruction gets called, so that on 
next run you can set up suitable breakpoint just before the crash, so 
that you can investigate state with last instructions in more detail.

When doing single stepping, it can help to write a debugger script that 
shows all the info you might be interested, and tell debugger to execute 
that on every debugger invocation (= on every step):
> lock file <debugger script>

To get normal debugger prompt output back, use:
> lock default


	- Eero