On 12/11/14 11:15, Richard Biener wrote:
Please find a better way to communicate possibly_undefined_stmt than enlarging struct loop. Like associating it with the niter bound we record (so you can also have more than one).
Unfortunately, the bounds get regenerated frequently, but the upper bound can only get reduced once, so the subsequent generations would lose the possibly_undefined_stmt.
I did originally try warning at the point where the upper bound gets reduced, but there were too many false positives. The two part solution I found gives the result I was looking for, but maybe there is another way?
I'm going to keep looking, but any suggestions would be appreciated. Andrew
