[psoc-2008] Week of July 7th, progress report

[Andrew Whitworth]

Progress on the new garbage collector is continuing at a regular pace,
although I appear to be moving at a much slower pace now then I
previously was. This is because the code is written, and the debugging
has started in earnest. I've already tracked down a number of bugs,
some that had to do with conceptual errors on my part (such as
sweeping the constant PMC pool, which shouldn't happen) and others
that appeared because of mistakes in my code. One such mistake I had
made previously was inverting the order of loops between different
versions of my allocator, which caused the sweep code to be moving in
the wrong direction. Another error involved my separate sweeping of
the string pools and the constant-sized pools: Since the string pools
are in the constant sized pools, I was sweeping them twice (and thus
freeing all strings prematurely). This has since been resolved.

With big obvious problems out of the way, I've moved to a phase of
much more difficult and tricky bugs to find. Visible progress will
appear slower, even if the amount of effort expended is the same. We
have entered the far less glamorous phase of software development.

The coming week involves, as vague as this may be, a continuance on my
bug hunting activities. I still have errors involving
prematurely-freed PMCs, and am working hard to isolate and eventually
resolve them. After I have resolved most visible bugs, I need to
reintegrate some of the code that I have removed for debugging
purposes: The system stack trace code (which was causing weird
segfault errors) and the Memory_Pool compacting code which was
mis-aligning data and b0rking string pointers (which, in turn, caused
more segfaults). After all this work, which will take at least a week
in the best case, there are a number of additional features that I
haven't even started work on yet, and would be dandy to have in my
final deliverable.

--Andrew Whitworth