A new overflow error in 2.42.6c

[Testlund]

I was away for a day hoping that I was going to run a nice evosim. When I came home this afternoon the program had crashed after 300000 cycles. I have attached the save here.

[EricL]

Testlund, your on a roll!

Bot 575 in this sim is attempting to make a virus.  The only issue is that its DNA is over 25000 base pairs long and the gene in question is 23503 base pairs in length!  The code multiplies this by 2 to get the number of cycles to wait before firing the virus, which results in an integer overflow of .vtimer.

I'm running out the door to go play in the sun so I haven't fixed this yet, but it will be fixed in 2.42.7.  Will probably make .vtimer a long.

Testlund, I'll post a buddy for you this evening (US west coast time).

I assume everyone is okay with a max DNA length of 32000?  There may be more bugs of this type coming out of the shadows as we see bots with DNA lengths approaching or attempting to exceed this...

[Testlund]

So that's what happend??   That was a quick evolve of the bots. Unfortunately I can't check it out for myself because all my aurosaves crash too. I think 2.42.7 will be awesome!  

[Numsgil]

I'd ideally like bots to have the possibility of up to 2^32 base pairs instead of 2^16, but it doesn't come up all that often

[EricL]

I'd ideally like bots to have the possibility of up to 2^32 base pairs instead of 2^16, but it doesn't come up all that often

Doing that would be a lot of work.  There are a lot of places in the code and in the DNA programming model where there are direct and indirect assumptions of being able to address DNA locations using an integer valued index and storing such indexes in the integer-valued DNA or mem array.  .vloc, .ploc and postive shotvals are direct indecies into the DNA for example.  .vtimer is indirectly a function of DNAlen and gets stored in the mem array.  A lot of the refvars (and sysvars (e.g. .myeye) similarly could in theory overflow if DNA was allowed to exceed 2^15.  Changing all this would be... a lot of work and amoung other things, I beleive necessitate moving from integer valued base pair and mem values to dwords.  Can you say win16?

It could be done, but IMHO, not high on the list for the VB fork...

[Numsgil]

It's done somewhat implicitly in the C++ version, so I guess I'm not going to worry about it too much.

[EricL]

Testlund, here is your buddy drop.  Has fixes for the long gene issue and the small kilobody veggy feeding thing.
Darwinbots 2.42.6d

[Testlund]

Kewl! Time to start a new evosim again. Hopefully it will last a little longer this time.