Apparently, this has been fixed upstream now. --
--- Begin Message ---On Sat Mar 10 19:45:35 2012, tonyc wrote: > Attached: regression test and possible fix. > > It may be that the code should be checking for magic earlier, but with > this change the test case above behaves as with 5.10.Applied as af89892ed and 05a1a0145d. And requested in cherrymaint. Tony
--- End Message ---
