[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
status (meeting?)
not sure if we're having a regular tuesday meeting, b/c of siggraph,
so here's a quick status report from me below. mic and i met up
monday, so unless someone else speaks up i might optimize the tuesday
trip to cambridge out of my schedule and work from home. let me know
if we are meeting. otherwise/regardless, cool developments.....
1. i have the remote wk database server/client working in most avail
configurations (irix/linux, local/remote), and intend to debug the
remaining failure cases i've seen pretty soon. this is new: this
stuff hasn't run since the last time rick worked on it i think. :)
getting it to go was not too hard tho.
2. i have the multiuser code working on linux, local and remote.
again, this is new, as above. plus i have it failing horribly but
in a way i at least abstractly understand on irix, and think i can
fix within a few more days.
note, this is a sim plugin, and one that really exercises quite a
lot of the generic threaded/server/client code layer. as such it
is an incredibly good regression testbed to work with. it is also
complicated enough i had to go through no less than 7 layers of
thread spawning (maybe 9 now that i'm counting) to get down to where
the bugs i'm seeing were... after i realized it was threading at
all! and not even threading grauitously! well maybe a little, i
talked to mic about it, and rwb may have gone overboardhyperactive a
bit but in the end everything he put in the framework is actually
useful (even indispensible) in the grand scheme. if we do end up
extending this further (like we hope to) it will be absolutely
necessary to have that machinery there.
3. i think i have a pretty good idea of how all the ndb stuff goes
together, and related stuff from there. read bukowski's phd thesis
if you ever want to get a start on it yourself, copy cached in
graphics:/home/luka/Papers/rwb-iwes.pdf. while that doc is not
half the story it's really one of the most well-written informative
theses i've seen in a while, and i'm working on getting the
remaining folklore into written form myself now. progressing well.
[must say: very few serious code hacks have been needed in this
latest development phase, mostly just figuring out how to pilot the
beast. i'm turning that "piloting/folklore" into docs now. plus
there are in fact a few needed patches. but on the whole, this is
some amazingly sophisticated and well-done code. rick bukowski
deserves some mad respect.]
4. i purified. i got scared and ran away. there's gonna be a lot of
cleanup to do before i can call this a stable system but at least
there's a precise way to assess what needs help once again now.
we're definitely not in any untenable position, and could still make
walkthru into a very nice stable high-quality package within a month
now i think. just keep the caffeine flowing.
5. i've done little else for the week that's demonstrable at this point.
but there are other things in the works of course. notably, some of
the above accomplishments clear dependencies in my table for dealing
with higher-level things. (e.g. once i have a regression case for
sims [via MU] i can attempt to depollute the sim support code, and
likewise the vis code that got too intimate with it via tapestry, and
then work on a sane vis codebase.)
all for now. i'm sure i'll have more in another hour as always. --pl