[Simulavr-devel] status report

[Theodore A. Roth]

Hi All,

Since the list has so quite I thought I'd send out a status report on what 
I've been doing.

For the most part, I've been writing regression test cases. There is now
35 test modules in the test_opcodes directory. The
regress/test_opcodes/TODO list says there are still 56 to write. This is
going to take a while to finish. Here's the results of the last test I 
ran:

  Ran 7943 tests in 269.880 seconds [29.432 tests/second].
    Number of Passing Tests: 7943
    Number of Failing Tests: 0

This was run on a dual PPro-180 system. I ran the tests on a PIII-550 and 
got about 72 tests/second. I'm sure I've missed some permutations in those 
tests, so I would say they are not 100% conclusive yet.

I did find and fixed another bug in the decoder.c today which would most
likely never have been found without the test system.

I rewrote some of the regression system to better encapsulate redundant 
checks and to hide the target interface in the base class. This makes 
writing test cases simpler but still tedious. Also, I ripped out the 
threading of the regression engine since a change I made to gdbserver.c 
(see next item) made it faster to run the regressions in a single thread.

While working on the latest cvs gdb, I found a bug in gdbserver.c. In 
fixing this bug, the gdb-simulavr interface was sped up quite a bit. A 
nice little bonus ;) As far as I'm concerned, gdbserver.c is ready for 
0.1.0. (Watch points have been deferred to 0.1.x).

I started to clean up the header files but it has been interrupted by 
regression stuff. What's the consensus on one big header file versus a 
header file for each C module? I'm leaning toward the latter, but am not 
sure what I want to do.

As for documentation, I haven't even touched it.

Things are progressing toward 0.1.0, but there's still quite a bit of work 
to do before the release.


Ted Roth