Most of the time, Argument Clinic just uses the line number of the stop line when it reports an error. This is not a huge problem, since an input block is normally pretty short, but it is slightly annoying when debugging, since you have to locate the faulty line in the input block.

In gh-107551, I'm reworking the error handling of Argument Clinic, which revealed an opportunity to get this fixed, or at least improved. That PR only introduces a crude workaround, though; it is not a permanent solution.