AVR-Chat

>> > All the programmers say that they can program and verify the chip
>> > successfully, and that all the fuses are set right. (including M103 
>> > fuse,
>> > CKOPT, WD, and BOD fuses)  Sometimes they fail to verify, or read wierd
>> > fuse values, but they are all in agreement at the moment that 
>> > everything's
>> > fine.
>>
>>And the OCDEN fuse ?
>>
>> > All three Jtag ices agree that they can load the program, but cannot 
>> > read
>> > the SRAM, registers, or anything else, and cannot run, or even
>> > single-step.
>>
>>See above.
>
> Enabled.
>
> Also wouldn't explain the total failure to run the program independently 
> of
> the ice.

True.
I just remembered I had a similar problem some time ago. I programmed the 
chip
but it didn't run. I verified the code, recompiled, programmed, verified - 
nothing
worked. After trying variants of this for a while, and pulling a decent 
amount of
hair, it turned out to be a problem with the code. Can't remember exactly 
what,
but the tools were actually working perfectly all the time.
As you can program and verify, and the fuses looks good, what else is left 
but
a code problem.
Try read out the code from the working board and program that on the other 
two.

Also, another trap is that setting the wrong CPU type with the STK-500 
programmer,
will NOT hold back programming or notify you in any way. Been there, tried 
that.

/Jesper