Bob,
When you hit ^C (once?) did you wait for the .hat file to be patched up after the signal was caught?
I didn't, this is a ghc on Mac bug -- ctrl-C doesn't work, you have to send it a kill -9 from elsewhere... I just forgot that. So is the likely cause that the file had not been flushed to disk?
It is more than just flushing an ordinary file buffer. Hat has its own fancy buffering system to minimise the number of updates to the trace file. Further, the trace has to be updated so that all expressions in mid-evaluation have "interrupted" as their result, and the trace must record the point of interruption. None of this can happen after a violent "kill -9". The program just stops dead. So does the trace. Hence the complaint from the trace-processing tools. Colin