java.io fix and speedup

[Tom Tromey]

>>>>> "Jeff" == Jeff Sturm <jsturm@one-point.com> writes:

Jeff> I have a local patch similar to Mark's that mmap's a
Jeff> FileInputStream.

That sounds interesting.  We'll need to keep the non-mmap
implementation, too, since mmap is a portability morass.

Doesn't the glibc stdio mmap in many/most situations?  If they do it
then it seems like it must be a performance win.  (I still haven't
looked at java.nio but my impression was that it was designed to allow
this sort of thing.)

Jeff> Given that getFilePointer can't really be trusted because it
Jeff> can't be performed atomically with any other operation, I don't
Jeff> see any harm in Mark's patch.

Thanks for opining; I agree.  Mark, go ahead and check it in.

Tom