DSIErrors with libraries

Hello,
Every now and then I am getting DSI errors with gl4es and Warp3dNova and am wondering if it is something in my code that is causing the error, or in the libraries themselves (so not my bug but rather a bug in the aforementioned libs).

For example, I get the following error:


Likewise, I have had simliar errors with gl4es.

Is there any way to debug these failures?

Thanks
Dave

@davec555

You need to look at assembler, you see what PowerPC register is used, and then look up the GPR register, and see what address is in it, it’s a DSI error so its data access error, can tell if on read or write / source or destination. So you get hints.

Now that might not be all you need to determine the problem. You might need a few printf to pin point the variable. The stack trace you did not provide tells you last function before the crash.

You can’t really be 100% sure if there is bug in your program or the library, just from DSI error, you kind need verify that provided correct values, than determine, if its bug in your program or in the library.

Developing on Windows / Linux is easy level.

Debug on AmigaOS4.1 / PPC is hard difficulty level.

Debugging on AmigaOS3.x / 68K you extremely hard difficulty level, you might not even notice the bug.

@LiveForIt Thanks for that. Its all very esoteric to me, so not really sure what is going on.

Here's another example (this time full crash log). Module pointed to in crash was the kernel.




So looking at this, I'm assuming register 8, which has a value of 0x00000007 but I don't know what that means.

Looks like the last call in my code was to draw_renderlist() and obove that is a call to ogles2.library.

Any help in learning how to do this is greatly appreciated.

Dave

@davec555

lwz r8,-16(r4) ; load word(int32) indexed, offset from r4 is -16

(half word is int16)

No it loads from GPR 4, so value is 0x4FC04CA0
so actaul crash is on 0x4FC04CA0 - 0x00000010

now you don't know what that is before you compare it to pointers you give to the function.

it loads from r8 is used as temp register, and its saved back into address of R9.

stw r8,-16(r9) ; store word (int32) into address r9 at offset -16

It looks like several offsets from same registers, so maybe it’s a struct that’s copied, can be a small array as well who knows.
not everything will make sense, but it often does give a hint.

@LiveForIt
Thanks. That makes reading that debug dump a little clearer.

So this part:
Quote:

So how do I determine what 0x4FC04CA points to? And what is the 0x00000010 (is that from the dar DataAddress)?

Thanks for helping out with this.

Dave

@davec555

hex 0x10 is 16 in decimal, You need to learn some computer science.

printf("ptr %08x\n", (uint32) ptr);

if value is the same as in register you know ptr is the same as address.

perhaps its corrupted, not set (random value), or someone has freed it, maybe memory is too small. wrong number in a array, a bad for/while loop.

Something like "ptr++"; can easily make a pointer go outside of range.

So finding the exact line before the crash is important, and perhaps the function name, due to inlining it might lie about function name in the stack trace.

printf(“%s:%s:%d\n”,__FILE__,__FUNCTION__,__LINE__);

Should help you find the exact line of the crash.

@LiveForIt
Quote:

Ha! I thought it was a number you pulled from somewhere else not address MINUS 0x10. I'm so dumb


...and THANKYOU for explaining.