
any time we call uncomment twice on some text, it probably won't have the desired result.  this happens in pascal and groff styles.

add more tests

speed it up.
  the slowness comes from executing lu-comment-extractor every time.
  and when we auto-detect, we throw away our results, only to get them again.


ctrl+c during detect doesn't always stop it.

byte offsets would be better than blocks in display of detection results


check if net tests are being skipped
  they are, erroneously.  why?
