Composition's poly expected degree mismatch about winterfell

[irakliyk]

Yeah - this is an unfortunate consequence of using pre-defined degrees vs. actual degrees. There are two ways around it:

1. If it is possible to insert random values somewhere in the column, the degree of the column would be guaranteed to be trace_length - 1. I've done something like this in one of the examples.
2. Another approach is to set the number of trace exceptions to 2 and then insert random values in all columns of the last row. This would also guarantee that degrees of all columns would be trace_length - 1.

It may also be possible to update the library in the future to rely on actual column degrees (instead of forcing them to always be trace_length - 1. Originally, I decided against doing this because I assumed that checking the degree would be pretty expensive. However, I've never actually benchmarked this - and it is possible that the extra expense is negligible in the grand scheme of things.

[4l0n50]

Thanks for your answer! Then maybe this is just about changing the error message. Right now it is not possible to accept columns with degree < trace_length-1 as part of the witness, and the message should indicate so.