CD32 NVRAM bug

[saimo]

While testing my (rewritten) NVRAM routines, I noticed an odd issue: if the NVRAM is written to starting from a certain address, the WinUAE NVRAM file grows beyond 1024 bytes. More precisely:
1. my tests write 6 bytes at arbirary NVRAM addresses between 0 and 1018;
2. if the address used is between 0 and 1008, the behaviour is normal;
3. starting from address 1009, the NVRAM file grows by address-1008 bytes.

Here is a picture that shows it visually (click to enlarge):



Moreover, the extra bytes get apparently random values: after restarting/resetting UAE and running the same test again, the values change.

If they are of any help, the archive attached here contains the test programs and the corresponding NVRAM files.

I've tested this using the public beta 31 and also the beta 0 2020.03.20 you gave me with the NVRAM debug log enabled.

[Toni Wilen]

Moved, I am 100% sure this has nothing to do with 4.5+ betas..

[saimo]

Sorry for posting it in the wrong place, then!

By the way, one thing I didn't mention (but you might have noticed it already): the problem is visible only when writing to the last 16-byte page. It looks like UAE always saves 16 bytes, starting for the intial write address.

[Toni Wilen]

I didn't test but it should be fixed now. Data write wraps around but file write didn't.

[saimo]

Cool, thanks!

[saimo]

I've tested beta32 now and I can confirm that the bug is gone
Thanks once again for the immediate fix!