Memory watch point on blitter writing to memory?

[mcgeezer]

Does anyone know if it is possible to set a memory watch point in WinUAE that is triggered when the blitter does a write to the memory?

It seems I can only get this to work with the CPU but the issue I have is that I have a rogue blitter operations somewhere causing memory corruption and I can't find it.

Any help appreciated,
Geezer

[roondar]

I haven't tested this, but perhaps you can use the 'write to address' breakpoint feature of the debugger and set it to trigger when the blitsize register is written?

That might work.

[chb]

also not tested myself, but winuae should support memwatches for all dma channels, see this thread:
http://eab.abime.net/showthread.php?t=77340

[Toni Wilen]

BLTD (in place of "CPU") works in this case. (Didn't you already ask this few month ago? Someone surely did)

[mcgeezer]

Originally Posted by Toni Wilen

BLTD (in place of "CPU") works in this case. (Didn't you already ask this few month ago? Someone surely did)

Thanks Toni.

I asked about logging Blitter operations...similar question I guess.

[mcgeezer]

Originally Posted by roondar

I haven't tested this, but perhaps you can use the 'write to address' breakpoint feature of the debugger and set it to trigger when the blitsize register is written?

That might work.

But that would cause every blitter operationto breakpoint... I just want to break when the blitter changes a memory location that I'm not expecting it to.

[Antiriad_UK]

Could you create a new CPU routine that runs periodically and it does a kind of check on the memory area (don't know if the corruption is a pattern you can check for)? If the routine finds something it doesn't like it could set a flag somewhere - then you could set a breakpoint on that flag.

[mcgeezer]

Originally Posted by Antiriad_UK

Could you create a new CPU routine that runs periodically and it does a kind of check on the memory area (don't know if the corruption is a pattern you can check for)? If the routine finds something it doesn't like it could set a flag somewhere - then you could set a breakpoint on that flag.

Believe me I've been knee deep in this.

Marking the start of every function and putting canaries all over the place.

The problem is so intermittent it is frustrating (it might do it 1/20 times).....it doesn't actually crash the game so I don't get a clean state when it happens.

[hooverphonique]

@Geezer: Are you using COPJMPs by any chance?

[mcgeezer]

Originally Posted by hooverphonique

@Geezer: Are you using COPJMPs by any chance?

In some places i do yes.

[ross]

Originally Posted by mcgeezer

In some places i do yes.

You have to be very carefully!
http://eab.abime.net/showpost.php?p=...&postcount=122

There is a situation, with the Copper waiting and the Blitter running, where a write with CPU to COPJMP update a blitter channel pointer instead of the copper address.

Why do you write to COPJMP? There is usually no reason.

[hooverphonique]

Originally Posted by ross

There is a situation, with the Copper waiting and the Blitter running, where a write with CPU to COPJMP update a blitter channel pointer instead of the copper address.

This is what I was hinting at, yes..

[mcgeezer]

Originally Posted by ross

You have to be very carefully!
http://eab.abime.net/showpost.php?p=...&postcount=122

There is a situation, with the Copper waiting and the Blitter running, where a write with CPU to COPJMP update a blitter channel pointer instead of the copper address.

Why do you write to COPJMP? There is usually no reason.

OK, I think I understand but the conditions for writing the COPJMP are not happening when the game corrupts.

In answer to the question, when Rygar enters the sanctuary scene and the bonus text is displayed I jump to another copper list to display different text colours.

I'm pretty sure this is just a case of I'm not managing game objects properly and I have a memory leak somewhere, I'll find it soon enough I'm sure.