Comment 19 for bug 1660142

Just checked the last bullet point above out of curiosity:
In fact without the typo fixed we initially detect *every* relative file link as "broken" (even if it's not).
However due to an (in principle) superfluous check we always looked at the "broken" link again just to see that it was not broken to start with (therefore no replacement was done).

In trunk we now correctly detect a valid relative file link and do not operate on it at all (therefore also the behavioral change described in bullet point 2 above should be fine).