+ /* Consider coding a repeat offset match.
+ *
+ * As a heuristic, we only consider the longest length of the
+ * longest repeat offset match. This does not, however,
+ * necessarily mean that we will never consider any other repeat
+ * offsets, because above we detect repeat offset matches that
+ * were found by the regular match-finder. Therefore, this
+ * special handling of the longest repeat-offset match is only
+ * helpful for coding a repeat offset match that was *not* found
+ * by the match-finder, e.g. due to being obscured by a less
+ * distant match that is at least as long.
+ *
+ * Note: an alternative, used in LZMA, is to consider every
+ * length of every repeat offset match. This is a more thorough
+ * search, and it makes it unnecessary to detect repeat offset
+ * matches that were found by the regular match-finder. But by
+ * my tests, for LZX the LZMA method slows down the compressor
+ * by ~10% and doesn't actually help the compression ratio too
+ * much.
+ *
+ * Also tested a compromise approach: consider every 3rd length
+ * of the longest repeat offset match. Still didn't seem quite
+ * worth it, though.
+ */