nine robots a-go-go

[statespacedev]

can celebrate a little bit:) after some intense macten hackn - solid progress
what we're getting right now is focused purely on letting the robots run free in utexas -
probly sacrificing some cis related stuff in utexas for that - can bring it back in later where it makes sense

[drforbin]

What changes did you make?

[statespacedev]

got a bit wild last night hahaha, will push a commit to utexas this weekend -
nutshell - changes to locking in gtkn, input, pause, and lock/unlock -
more soon amigo, will clean it up and write it up a bit

[statespacedev]

alright, bit more progress, and jotting down a post around learnings - here's copypasta of some of that, and 19b1253

• there are some crashes to deal with
• what's the most effective way for a beginner to approach this situation?
• one possibility is to simplify things as much as possible

replace the players with robots. no drops, hangups, or sudden exits from the game. where it makes sense, limit or deactivate code for handling connectivity and continuity, of which there is plenty from compuserve for their dialup customers. then base everything around a smooth and friendly stream of commands from a limited vocabulary. 'move' and 'tell' primarily. some 'list' and 'time'. a bit of randomness around the exact timing of the commands, the 'move' direction, the 'tell' messages.

at the moment things work well enough if the robots don't use the 'tell' command. this suggests, focus on what's happening inside the 'tell' command and try to make things work well enough with it active as well. things have been simplified to the point where they can work ok, but it's on the edge. in fact, the 'tell' command can be active as long as there's not 'many' robots. it's both the 'tell' commmand and many robots that push it over the edge. this suggests that 'system load' is involved, whether via number of robots, frequency at which they send commands, types of commands, etc.

meanwhile, the locking code is fundamental, so also simplify there and make it as stripped down and well understood as possible. if locking can possibly work without a particular line of code, comment it out. if not forced to bring it back, leave it out.

[drforbin]

Noah,
I had all kinds of problems with tell command....I fixed this in devel

[statespacedev]

yep understood merlyn - and your devel work is utexas starting point - actually have a feel there's trouble at least partly on the tops10 side - have now seen full on tops10 crashes by speeding up the robots just a little bit - they can easily put real stress on tops10 and the 'hardware':)

[statespacedev]

yeah, the more am thinking about the 'feel' that's building up over many weeks of stress testing with robots...
nutshell - being 'gentle' with the game / tops10 / pdp10 makes a huge difference
is what it is - we're gonna be building the robustness more and more - that's our 'job'
but these were temperamental beasts back in the day, no?;)
fwiw fyi - this one is big - we're next level hombres;) 6a39507

[drforbin]

I guess Noah, your at the for front of all this. I remember tops10 as being very robust.
dual processor kl10's routinely handled over 100 jobs for years without a reboot.

[statespacedev]

beautiful, thanks merlyn - true
that thought is in the brain too
the legend is based on reality - we know that
then this fourteen year old kid knew about crashes playing decwar - san marcos tx circa 83/84
this is the fun part really... - we're getting down into the heart;)

[statespacedev]

lookn good;)