Darwinbots Forum
Code center => Bugs and fixes => Topic started by: PurpleYouko on May 04, 2006, 02:42:02 PM
-
OK then (cracks knuckles) I have just downloaded V2.43b for a good old test run.
Here are a list of the things that I think need fixing or modifying.
- A tie bug that Nums managed to squash back in an early version of 2.37 has rea-appeared. In a simple tie-feeder (HDV4) against veggie sim I notice that ties between very distant objects often spring into existence and then disappear again. I don't remember how we fixed this previously. maybe Num can remember.
- Bots appear to be able to quite easily break free of a tie by accelerating away from the bot they are tied to. Maybe ties need to be stronger to prevent this.
- Mass (in the console window) appears to be reading back a value of zero on all the bots that I have checked so far
- When physics is set to "thick fluid" the bots just sit there and vibrate back and forth (test performed using "Anon Terifica") giving a kind of overlapped double image. I finally managed to get this to reset by going to "thin liquid" settings.
- It is not possible to deselect one of the liquid or solid options once selected. Also the advanced controls don't appear to make one jot of difference once you have selected one of the main options.
- I can no longer see gene activations in the console window when I advance one cycle. This is just about the most necessary piece of the code in my opinion. I can't develop or debug a robot without this.
- Certain robots can be sitting there pounding a veggie with shots and growing in size then suddenly they die. "Destinatus Preliator" does this all the time. I can't tell why because it impossible to see which genes are activating. (see point above)
I just followed a single specimin around the screen for several minutes while he searched for food. His energy was dropping very slowly as DP is designed to manage his energy very well. He spotted a veggie, rushed in and began to shoot it. his body began to sewll immediately but he died within 20 cycles.
It seems to cost more to fire a shot than the shot gives back DP was losing more than 60 energy per cycle, some of which is going to body but that doesn't explain it all. - after a couple of thousand cycles I got an overflow and it saved an error.sim file (included in this post.
That's about it for now.
In a nutshell, I can't really get any of my bots to work properly and I can't diagnose what is wrong with them.
The have no discernable control over the physics functions and something is dragging down the energy of my shot-bots.
I am willing to help debug 2.43b but as for being a stable and usable version, I think it needs a lot of work before we can call it that.
Oh for christ sake!!
It won't let me upload a .sim file. I have renamed it to "error.txt" so you will have to change it back before loading it up.
-
I'll investigate and respond in detail later. For now, could you zip the sim file and reattach it? Roundtripping through a txt file is corrupting it I think. Thanks.
-
OK here it is as a zip.
-
OK here it is as a zip.
It loads and runs fine for me adn ends up with nothign but veggies. Probably the bot that hit the crash is long gone. Can you maybe zip up and attach the original sim you started with so I can start from there? Thanks.
Also, are you really saying the gene activations dialog doesn't work for you or are you simply lementing that you want it to be part of the robot properties or console dialog? Gene actvations has been working for months....
-
I am not referring to the gene activatins window. I am referring to the console window which allows you to step through the game cycle by cycle and shows which gene is activated during each cycle as you do so.
It outputs a long scrollable list to the console window in every version that there has ever been prior to this one.
here is what it looks like in 2.37.6 and all previous versions.
[a href=\"http://img344.imageshack.us/my.php?image=2376simulation6jy.jpg\" target=\"_blank\"]
-
here is an example of the tie problem.
To the right of the picture you will see one of my Carnatus Orbis bots tied to (and apparently feeding from) absolutely nothing.
(http://img60.imageshack.us/img60/8616/carnatusorbiswithatietonothing.th.jpg) (http://img60.imageshack.us/my.php?image=carnatusorbiswithatietonothing.jpg)
-
Seems to me that a tie isn't getting deleted when a bot dies for some reason.
I don't think it's the previous tie bug we had. That was very specific and there should be several paragraphs in the code detailing what it was and how I squished it.
-
It does appear a little bit different than before.
Impossibly long ties appearing and disappearing in the course of a single cycle. Others tying to invisible objects.
Definitely sounds like ties not being deleted when a bot dies in some cases but in others the ties just appear.
In my tie feeder sims, I see loads of ties forming from a tie feeder to a distant veggie then disappearing just as quickly. It looks as if the robot is attempting to make the tie and the program is allowing it to form even though it exceeds the maximum length. on the next cycle it is deleted again.
-
I'm getting that too.
Here you can see two Devincio Solos tieing to nothing
(http://img182.imageshack.us/img182/276/tiestonothing6sq.th.jpg) (http://img182.imageshack.us/my.php?image=tiestonothing6sq.jpg)
Here you can see Endy's Alga Tier forming a tie to a distant object
(http://img222.imageshack.us/img222/4286/longrangetie5ek.th.jpg) (http://img222.imageshack.us/my.php?image=longrangetie5ek.jpg)
Some more long-range ties
(http://img243.imageshack.us/img243/5508/longrangetie26ta.th.jpg) (http://img243.imageshack.us/my.php?image=longrangetie26ta.jpg)
Antibirthtie genes also sometimes don't work, although the ties are so weak that the bots can just pull them apart anyway
-
I am not referring to the gene activatins window. I am referring to the console window which allows you to step through the game cycle by cycle and shows which gene is activated during each cycle as you do so.
It outputs a long scrollable list to the console window in every version that there has ever been prior to this one.
Wow, I had no idea.... This has never worked in 2.4, the code was commented out. I didn't even know it was missing! The changes that were made prior to my coming on line regarding DNA execution necessitated my implementing it a little differently (the GA array is now part of the robot structure, but is only calculated for bots which have the focus or an open console window) but it is completed and will be in 2.42.4. See #3 here (http://www.darwinbots.com/Forum/index.php?showtopic=1309).
Question 1: Should the gene activations text be printed in the console window only when the single cycle button is pressed or also when the play button is pressed in a bot's console window? I assume they should not be displayed when the play button is pressed on the main MDI window.
Question 2: Should the activations be printed in all open console windows or only in the one in which the cycle button is pressed?
Here is the settings file from 2.42. Most of the stuff is absolute default. There aren't any weird settings or anything. I put it in as a zip because the forum probably won't allow it otherwise.
Um, sorry, but you could zip a sim instead of a settings file? I don't have the bots and the install locations are different (your bots on on your D drive) so sim would be much appreciated. Thanks.
-
Wouldn't it just be quicker and easier for you to just grab the bots from the bestiary and run them yourself?
As I said, I am using more or less default settings anyway.
I guess I cold set up the sim again, start it going and then save it but I can't do that till Monday now.
Question 1: Should the gene activations text be printed in the console window only when the single cycle button is pressed or also when the play button is pressed in a bot's console window? I assume they should not be displayed when the play button is pressed on the main MDI window.
In 2.37.6 they are output to the console window when the play button (the one in the console window) is pressed. This can often be very useful so I would say that it should continue to work that way.
Doing this results in a much slower frame rate than normal execution so it is a great help for debugging.
The normal play button should not do this. In fact I'm not even sure you should be able to activate the normal play button while the console window is open.
Question 2: Should the activations be printed in all open console windows or only in the one in which the cycle button is pressed?
I would say for all open console windows.
Older versions only had ONE console window but the advent of multiple ones could help immensely with debugging, particularly when watching comunications between elements of an MB or during a close up of a combat event.
-
Wouldn't it just be quicker and easier for you to just grab the bots from the bestiary and run them yourself?
As I said, I am using more or less default settings anyway.
No, it would not. But I did try. I'm over half a million cycles now (running in VB) using the default settings with Alga_Minimalis and Destinatus Preliator. I am unable to reproduce your crash. Without a sim file from you with your exact settings and bots, I cannot pursue this further. It could be the bot color you choose for all I know.