Applying either of the two solutions described above would fix the errors. For example, changing the memory model to large would place the 192 bytes of variable data in XDATA. Similarly, we could manually place the variables in another memory segments such as IDATA, PDATA, or XDATA.
BEWARE, using the large model will make evrything slow, other ,mysteries' may occur. I have seen timing issues galore from large model users.
were you working for me the statement would be "use the large model and you are fired".
a good rule of thumb would be arrays and buffers in XDATA, everything else in data