Darwinbots Forum

Code center => Bugs and fixes => Topic started by: Shen on April 12, 2005, 06:09:31 AM

Title: 2.36.3 Bugs
Post by: Shen on April 12, 2005, 06:09:31 AM
The left wall is solid for some bots. Cant figure out why yet but some bots are fine and others act as though its non torroidal.



Minor niggles.

Changing the field size to a smaller size messes up the starting positions of bots, they all get put in the top left corner and other wierdness.

Respawned alga doesnt obey the starting position defined in the species tab in simulation settings.

Waste Threshold isnt saved in the settings. :)
Title: 2.36.3 Bugs
Post by: PurpleYouko on April 12, 2005, 09:04:29 AM
Quote
The left wall is solid for some bots. Cant figure out why yet but some bots are fine and others act as though its non torroidal.

I noticed that shortly after sending the code to Num. Working on it today.

Quote
Minor niggles.

Changing the field size to a smaller size messes up the starting positions of bots, they all get put in the top left corner and other wierdness.

Respawned alga doesnt obey the starting position defined in the species tab in simulation settings.

Don't know anything about start positions. I have never worked on that code.

Quote
Waste Threshold isnt saved in the settings.
Nor are a lot of other things.  :rolleyes:
We need to have a major overhaul of the save routines but in all honesty I think we have been putting it off until all the changes have been made and are otherwise working. It is highly difficult to make save and load routines backward compatable.

Thanks for the comments.  :D
Title: 2.36.3 Bugs
Post by: shvarz on April 13, 2005, 12:20:55 AM
I tried the unofficial 2.36.3PY version - it is still unstable for me.  Crashed within 15 minutes (the Overflow error).  The 2.36.1 version can run for 10 hours without a problem.
Title: 2.36.3 Bugs
Post by: PurpleYouko on April 13, 2005, 09:33:22 AM
Quote
I tried the unofficial 2.36.3PY version - it is still unstable for me.  Crashed within 15 minutes (the Overflow error).  The 2.36.1 version can run for 10 hours without a problem.
DAMMIT!  :angry:

Send me your settings file please so I can try out exactly what you did.

I couldn't make the thing crash at all. Ran for several million cycles with no errors.
Title: 2.36.3 Bugs
Post by: AZPaul on April 13, 2005, 12:51:59 PM
Hey Peoples,

Kill a bug over here, grow a bug over there. Happens with complex code.

DB 2.36.3PY

1.  Bot velocity exceeds maxvel.
2.  Inserted bots often constrained to slow velocity.
3.  Velocity values of 90+, inconsistantly.
4.  Velocity values become more inconsistant the longer the sim runs.

And so as to not leave the impression that the "Cambrian Explosion" was all nice and safe and productive with abundant new life in protected new niches exercising new freedoms to try copious new body plans rapidly evolving while happily embarking in anticipation on vast new adventures across the dry lands...there were four mass extinction events during the period.  

I still think this popularized view (from within the discipline itself) of an "overnight explosion" in the creation of abundant new lifeforms is a wee bit melodramatic considering the facts. An example of Dawkin's "bad scientific poetry" maybe?

See, y'all.

-P
Title: 2.36.3 Bugs
Post by: PurpleYouko on April 13, 2005, 02:05:12 PM
Quote
DB 2.36.3PY

1. Bot velocity exceeds maxvel.
2. Inserted bots often constrained to slow velocity.
3. Velocity values of 90+, inconsistantly.
4. Velocity values become more inconsistant the longer the sim runs.

This is completely Nucking Futs  :(

I only changed one bloody procedure. And that was to revert back to an earlier version of that procedure. A version that, incidentally, has worked fine for over 2 years before V2.36.1

What the hell is going on here?
Title: 2.36.3 Bugs
Post by: PurpleYouko on April 13, 2005, 02:42:20 PM
What robots are you using?

I can't get any of them to go faster than maxvel  :blink:
Title: 2.36.3 Bugs
Post by: PurpleYouko on April 13, 2005, 04:13:48 PM
I just tried making a robot deliberately to go as fast as it possibly can. It spends down its body points till they reach 1.
It can still only reach a top speed of 59.9. I even set up a capture event in the code to trap any bot that exceeds 60 either for .maxvel or .vel

Didn't happen in over 2 hours.  :unsure:
Title: 2.36.3 Bugs
Post by: Numsgil on April 13, 2005, 05:09:25 PM
Paul, were you using walls?  I've gotten walls to go really fast, and impart that speed to other bots.
Title: 2.36.3 Bugs
Post by: shvarz on April 13, 2005, 06:08:14 PM
Nums, this sounds like you were trying to achieve that effect :)

Crazy-crazy Nums :)
Title: 2.36.3 Bugs
Post by: Numsgil on April 13, 2005, 07:00:51 PM
Actually I have no idea why walls are uranium balls waiting to explode.
Title: 2.36.3 Bugs
Post by: AZPaul on April 13, 2005, 11:59:01 PM
Quote
What robots are you using?

I can't get any of them to go faster than maxvel 


Wouldn't be at all surprised if somehow my DNA code morphs into something really strange. I won't hold my breath though.

I followed a Delta on here journey after inserting her into my sim. She killed a number of Alphas then, while chasing one she seemed to drag. Through the console I learned that, though the Alpha she was chasing was showing .vel 81 Delta was suddenly reading .refvel 34. I noticed this because Delta's eye values in eye4,5,6 had dropped from 43 to 32 in one cycle.  Looking at the other Delta in the sim for verification is where I found she was exceeding her (too low) .maxvel.  My experience is that Delta should maintain maxvel of 57. Suddenly having a maxvel of 18 (while .vel showed velocity of 19) with less than 600 body and no shell was a bit of a surprise.

I've attached my settings file and my two Bots in a zip folder. I'm trying to determine where a good threshold for a stable population is in a small #2 sim. This is preliminary to a #5 sim for my testbed. I let the Alpha population get up to about 100+ before inserting a couple of Deltas. After about 4-5,000 cycles you should see the issue.

Unless, of course, I have queertrons in my PC which are causing the problem and thus is not the fault of the application. I do keep my PC pretty clean though and have not experienced any issues with my other apps.
Title: 2.36.3 Bugs
Post by: Numsgil on April 14, 2005, 12:04:34 AM
Sounds like a massive but subtle bug somewhere in the program.  This is going to be a pain to find and fix.
Title: 2.36.3 Bugs
Post by: PurpleYouko on April 14, 2005, 09:43:45 AM
Isn't it though?

I reviewed all the code that controls maxvel yesterday and I can see no issues with it. All my tests worked exactly as predicted. The only way I can see the velocity getting bigger than 60 is if body has become negative somehow.
That should not be possible as the robot should be dead at this point.

I will check it out and see if I can reproduce this error.
Title: 2.36.3 Bugs
Post by: PurpleYouko on April 14, 2005, 09:48:03 AM
Can anyone download that zip OK?

It downloaded OK but unzipped into a folder containing one file full of garbled, unusable nonsense.

No robot files
No settings file
Title: 2.36.3 Bugs
Post by: shvarz on April 14, 2005, 11:18:05 AM
I see two robot files and one settings file.
Try d/l again.
Title: 2.36.3 Bugs
Post by: AZPaul on April 14, 2005, 12:50:07 PM
Quote
Can anyone download that zip OK?

It downloaded OK but unzipped into a folder containing one file full of garbled, unusable nonsense.

No robot files
No settings file

I just downloaded and got 0 files myself. Says file is corrupted. Strange, Shvarz got it.

BTY, I'm using Window XP Pro

Well, since 2.36.4 is out I'll try that first before going any further.

-P
Title: 2.36.3 Bugs
Post by: PurpleYouko on April 14, 2005, 01:21:43 PM
Quote
Well, since 2.36.4 is out I'll try that first before going any further.
There are still some bugs in that one too.

I just sent Num the code that I fixed yesterday with Shvarz's help.

My bits are already fixed in 2.36.3PY but are not incorporated into 2.36.4 yet. They won't effect normal sims anyway. I always rely on Shvarz to push the envelope and find some way to kill the program.