[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]

TextySnarfXcvr progress

A test run Sep 25 03:06 produced (in part) the following:

> Test of Rcv side of bare SnarfXcvr
> Shepherd@23(5,
>   Sheep@26(6,
>     NULL,
>     NULL),
>   Sheep@28(7,
>     NULL,
>     NULL))

That's the reception of a rather simple flock (one shepherd, pointing to two
sheep, no outpointers, no cycles).  The shepherd stub was transmuted into
a shepherd by the receive process, and all was hunky-dory.  This part of the
test has worked for most of the week.

The next part is more interesting:

> Test of Rcv side of bare SnarfXcvr
> [additional diagnostic output removed for clarity]
>   Sheep@53(2,
>     Angora@55(8,
>       NULL,
>       NULL,
>       12.34,5,-6,7,-8,(ab,d),'H&S C string'),
>     NULL),
>   Sheep@57(3,
>     Shepherd@5(5,
>       Sheep@6(6,
>         NULL,
>         NULL),
>       Sheep@7(7,
>         NULL,
>         NULL)),
>     Shepherd@50(recursive)))

This is the first successful migration of a full-blown flock by reception
of its shepherd (formerly @1) into a shepherd stub (@50).

 - The Angora (i.e. lots of hair) sheep tests the reception of all the
   currently available data types.  (All are perfect except IEEEDoubleVar,
   which can currently cause roundoff error in texty protocols.)

 - The Sheep @53(formerly @3) points to two shepherds:

    - The first was already incarnate.  The bogus ViewHandler (= SnarfPacker)
      returned a pointer to the actual object, to which the sheep now points.
      (Had it been stubbed, the ViewHandler would have returned a pointer to
       the stub, and had it been disk-only, the ViewHandler would have created
       a stub and returned a pointer to that.)

    - The second is the sheep's own shepherd @50(formerly @2), which forms a
      cycle.  Note that the sheep now points to it.

This exercises virtually all of the receive side code.

In the process, I identified problems in TextyCommXcvrs, Recipies, and the
Become mechanism.  Some of these have been fixed, and others temporarily
worked around.

Over the next several days I'll be pulling out the bailing wire and
scaffolding, propagating the changes into StubbleForm, Tofu, and down the
Snarf side and up the Comm side of the fork.  This will require some
coordination with other staff.

In case I don't make the meeting this afternoon, rest assured that VERY
good things are happening.