super slow crash when attacking with 80 Kalonian Hydras

[AlexFolland]

I made a ton of Kalonian Hydra s for the lulz using several Followed Footsteps and Paradox Haze s. When I attacked with 80 of them, Forge slowly and painfully tried to double all the counters on each one 80 times, but with replacement effects from 2 Doubling Season s. It was so slow and painful that I'm still sitting here looking at a blank crash report screen. I realize the resulting numbers of +1/+1 counters will be astronomical. Maybe there should be some sort of internal number check to prevent a crash.

Edit: Yup, the whole JRE crashed. I couldn't get a crash report.

Here's a fun screenshot from before the attack:

[friarsol]

I'm not sure. if you are that interested in trying to do that you probably deserve to have the game crash on you, right?

[AlexFolland]

Well, no, that's not how software development works. Software should have some sort of method to prevent itself from crashing. What I did is legal in MTG and even happens with only a few Kalonian Hydra s.

Edit: That's a pretty offensive suggestion, too. Nobody "deserves" to have software crash on them. I am submitting a bug report in order to help the project. I was messing around with power 9 in this game to reproduce it quickly which might make you feel like I "deserve" it, but I have a modern-legal deck with Kalonian Hydra s and Archangel of Thune s that has caused the same crash in regular play, causing a lot of lost progress at the time. I was simply having fun while trying to reproduce the crash and realized I should probably report it to help the project.

[friarsol]

Sorry you missed the joke.

[AlexFolland]

Half of a joke transaction is delivery, which includes appropriate timing and setting. I'm not sure my understanding is what missed the mark.

Do you have any commentary on the actual bug? I'm interested particularly in why the doubling seems to be a little messed up anyway. It somehow got a bunch of the Kalonian Hydra s to 2500/2500 before it crashed, even though that is between the correct 2048 and 4096, so the doubling should never return that number. What caused that? Could this apparent extra pseudo-rounding be causing unexpected problems? I would have thought the crash would come when the numbers got much bigger than only 2500. Maybe it was from the number of creatures that were necessary to process in one attack, or maybe all the replacement effects from the Doubling Season s.

[Xyx]

Note the tiny scrollbar to the right of my battlefield.

Step 1: Starfield of Nyx.
Step 2: Parallel Lives.
Step 3: Followed Footsteps on Parallel Lives.
Step 4: Watch the universe implode.

OK, so the universe doesn't actually implode, but three upkeeps later things slowed down considerably, so I left Forge running for half an hour, and when I came back it was still pooping out tokens at a rate of ~1 per second.

I don't know if anything can be done about that, or even if it should. It's deliberately pushing the limitations of the framework. This was after three upkeeps. What happens after four upkeeps... I'm not sure that can even be written in scientific notation, let alone handled by a java machine.

[Marek14]

On first upkeep, you should get 2 tokens (2^1). On second, eight (2^(1+2)). On third, 2048 (2^(1+2+8)).

On the fourth upkeep, you should get 2^(1+2+8+2048) tokens. This is about 6.6 x 10^619. Scientific notation won't break down until the fifth upkeep.

If every of those tokens occupied a single byte in your computer, you'd need 6.0 x 10^607 1 TB disks to keep them.

A portable 1 TB hard disk has a mass of about 300 g, so this mass of disks would weigh about 1.8 x 10^607 kg.

The mass of observable universe is around 3 x 10^52 kg, so the disks would be about 6.0 x 10^554 times heavier than the universe.

If you won't model the individual tokens, but only their NUMBER, then these calculations will hold for the FIFTH upkeep where you will need this much storage (and more) just to store the correct number of tokens.

[Xyx]

The mass of observable universe is around 3 x 10^52 kg, so the disks would be about 6.0 x 10^554 times heavier than the universe.

Hah, so maybe the universe would actually implode!

We still expect Forge to handle that gracefully, of course.

[Fizanko]

Maybe as some kind of safety belt, once the player (or the AI) have an amount of X cards on his side of the battlefield (what X could be, i have no idea as i never tested those kind of token multiplying decks to check how much my system can take before a crash), there should be a pause and message from Forge stating that the program is reaching its current limits and that if you continue to play the current game, crashes are to be expected.

Followed with a "do you want to really continue YES/NO" so if the player wasn't on purpose doing that (or had no idea the AI was using a deck resulting in such behaviour), he would still have a chance to exit the current broken game without having to deal with his system ressource being eaten and Forge crashing.