hi,

hmm, it seems that Atmel is not solitary in undocumented "features".
I say more, Atmel is even "better": they have documented such things which if normal people is looking on, then he drops the idea to use such "product". For example:

"In revisions “S” and earlier of the AT89S8252 and AT89LS8252 devices, inadvertent writes to either the Flash program memory or the data EEPROM memory can occur. An inadvertent write is a possibility in the following two scenarios.

Case A Conditions
1. A MOVX instruction is used to write to the internal EEPROM.
2. P2.4 = 0 (driven by external hardware)
    P2.5 = 1
3. The RST pin is asserted high during execution of any MOVX instruction that writes to the EEPROM, while VCC is on.
In this case, an inadvertent write to the EEPROM address location corresponding to the value at Port 1 and pins P2.0 - P2.5 can occur.

Case B Conditions
1. A MOVX instruction is used to write to the internal EEPROM.
2. P2.4 = 0 (driven by external hardware)
    P2.5 = 0 (driven by external hardware)
3. The RST pin is asserted high during execution of any MOVX instruction that writes to the EEPROM, while VCC is on.
In this case, an inadvertent write to the Flash memory address location corresponding to the value at Port 1 and pins P2.0 - P2.5 can occur."

or next:

"The Atmel AT89S microcontrollers use three SPI lines to implement in-system programming (ISP). When the ISP programmer asserts the RESET line, the target AT89S microcontroller is supposed to tri-state all I/O pins including the SPI lines and then wait for commands from the programmer.
Experiments using AT89S devices have shown that if the user code writes to any of the ISP I/O pins (MOSI, MISO, SCK) within approximately 500 ms of the user code coming out of RESET condition, any subsequent serial programming operation fails. The problem is then that the only way to re-program the device successfully is to physically remove it from the target board and erase it in a parallel programmer."

Do you like it?

Regards,
Oleg