From wdiff2.c:

      /* TRANSLATORS: This and the next string are one message. */
      fprintf (stderr, _("Launching `%s"), MDIFF_PROGRAM);
      for (counter = 1; counter < arguments; counter++)
fprintf (stderr, " %s", argument_list[counter]);
      fprintf (stderr, _("'\n"));

The above code makes it unnecessarily hard for translators. "This and the next string" should be merged into a single string. Here's an idea.

    const char *launching_msg = _("Launching `%s'");
    const char *percent = strchr (launching_msg, '%');
    fprintf (stderr, "%*s", (int)(percent - launching_msg),
launching_msg);
    for (counter = 1; counter < arguments; counter++)
      fprintf (stderr, " %s", argument_list[counter]);
    fprintf (stderr, "%s", percent + 2);

This puts the burden onto the programmer instead of the translators. Sure, the code is more difficult to understand, but the translators will find it easier to translate the complete sentence instead of having to translate two separate pieces.

Or if you don't like the %s placeholder, you could also take another character that is unlikely to occur in any translation. Or you just tell the translators to keep the %s as-is and not add any modifiers like %00000002s.