Corrected Status Report

[Tom Droege <droege@wwa.com>]

Corrected Status Report - Or The Danger of Testing on Binary Boundaries

Yep, the title says it.  The memory was not really working.  But it is now.
I was testing the memory board with an 8 Meg stick stuck into a socket
designed for a 16 Meg stick.  I was testing with a random block 2K long.
Everything looked just fine until I thought to write the address into a
block of 4 bytes instead of random numbers.  Then everything was really
strange. 
Usually random numbers will kill something that all other tests pass.  But
you still have to run the routine tests or else you get bit by something
like the above.   

On a new device, there is no real way to distinguish between a read address
and a write address failure.  Well, you have to be really devious.  A lot
of turning power off to clear the memory.  It turns out it was
neither/both.  It was writing and reading from memory that was not there.
It took me a whole day to recognize this.  

OK, there is a reason.  I was using the prototype memory board to check
against.  It was tested with the prototype camera with 8 Mbyte sticks.
Apparently Merle modified it without me knowing to expect either 16 or 32
Mbyte sticks instead of the original 8 or 16 Mbyte sticks.   Thanks Merle.
Actually he probably told me that he did it.  Then he took off out of town
at the time I decided to run tests.  Merle is always "testing" me this way.
 I guess I am still passing or he would have dumped me by now.  

The new memory board works, but fails when it warms up.  Sigh!  There is no
end of work in sight.   At least this is a "normal" type of problem.

Tom Droege