Re: [hatari-devel] fixed a crash when restoring a snapshot using DSP

[Laurent Sallafranque <laurent.sallafranque@xxxxxxx>]

Just to be sure everything is OK, at the end of DSP_Reset, there's  
save_cycles = 0;
We must be sure that this value is not too high (I don't know why ;), or 
the DSP may execute a lot of instructions after restoring the snapshot 
(this value is peharps saved into the snapshot too)

In dsp_core_reset, there's a call to dsp56k_init_cpu(void)

void dsp56k_init_cpu(void)
{
    dsp56k_disasm_init();
    isDsp_in_disasm_mode = false;
    start_time = SDL_GetTicks();
    num_inst = 0;
}

Which inits a few things, the most import is the disasm_init pointer 
which initiates the disasembler pointer to dsp_core if I remenber well 
(for DSP traces and disasm).
This may crash hatari if we try to debug DSP code after a snapshot 
reload and this variable is not set, I haven't verified.

Laurent



Le 03/01/2015 18:38, Nicolas Pomarède a écrit :
> Le 03/01/2015 17:23, Nicolas Pomarède a écrit :
> > So, latest changes prevented the crash, but it was not correct, I will
> > change it do not call DSP_Init after the snapshot is loaded.
>
> This seems to be OK now.
>
>  - DSP_Init is only called at start and doesn't modify bDspEnabled, it 
> only inits the DSP (DSP_Init() is now called in all mode, but we don't 
> want to enable it for STF for example)
>
>  - I added DSP_Enable / DSP_Disable to really enable/disable DSP 
> emulation. These functions should be called when changing config 
> instead of calling DSP_Init/DSP_UnInit.
>
>  - I also remove the call to DSP_Reset when restoring the snapshot ; 
> from what I can see it reset some variables in dsp_core, but as 
> dsp_core is loaded right after, the reset seems useless.
>
> Tested with GemPlay, I can now reload a snapshot and still use the DSP 
> and its program (but maybe some corner cases are remaining)
>
> Nicolas