MArcomage

What it was? First time i think it was part of my dream or somthing like that - i was really sleepy.
Incognito - turn 45 - wheat farm
Me - turn 46 - dark templar (discard some cards)
??? - turn 47 - all cards, discarded on turn 46 has changed!
Incognito - turn 48 - Citadel of Myr

http://arcomage.netvor.sk/?location=Replays_details&CurrentReplay=182054&PlayerView=1&Turn=45

it was not important in that game, i'm loose anyway. But i was really surprised, when noticed it.

Wow, yeah, I see what you mean. Turn 46 behaves as it should, and then turn 47 seems not to have been prompted by anything. It refreshed those spots that had just been replaced and did not generate a new Discard list during a time that nothing was played. I'm going to look at this more closely and see if I can figure it out.

Well, it's not performing the action twice. If so, Shifting mass would have remained on hand after the first go. Unless it's performing the action twice based off of the initial calculations. Looking at the card code now.

Card code looks legit. It makes one pass of each position, looking first at your card, replacing it if needed and flagging it New, then the opponent's card, replacing it if needed and flagging it New. Repeat for slot 2 3 4 5 6 7 8.

No idea. There is no reason it should do this twice, and it doesn't store a list of valid positions to replace, so I don't know why Shifting mass would have been negated with the rest. Anyone know of another replay where Dark templar was used so we can compare?

If I had to guess, it probably occurred somewhere in the turn order code rather than the card effect itself.

I just realized that Shifting mass wasn't a card drawn to replace a discard but the regular draw after Dark templar was played, that card having been played from slot 2. Interesting.

This looks interesting. I'll look into it.

Is MArcomage begining to be sentient? Is this the begining of Skynet?

Who knows... can be :D

I was unable to replicate this bug. The main problem is that the replay indicates contradictory facts:

- the current status indicates that the mysterious turn belongs to iNcoGNito

- the changes indicates that it is still SVirt's turn because the damage done by Dark templar to iNcoGNito was not reset yet

- the effect of the mysterious turn would suggest that iNcoGNito played a Dark templar (all cards with total cost below 10 were discarded from both players hands)

- iNcoGNito could not have played Dark templar, because he had no Dark templar on hand turn before

- the Dark templar played presumably by iNcoGNito did damage to iNcoGNito instead of SVirt

- no player got any production that turn

- no player payed any cost for the card played during the mysterious turn

I tested the cards and they seems to be fine. The error must be somewhere else, however there are no errors in the error log that would indicate that something failed. I don't know... The information I extracted from the replay only confused me more @_@. It would help much if iNcoGNito would remember how the mysterious turn looked like before playing.

Could it be because of a double click?
usually when you double click "play" (and your browser is slow) there is a message that says "action allowed only in your turn". it this situation it seems that the limitation didn't work and the player had played dark templar two times in a turn (maybe the second dark templar was not the same but it was from the normal draw)

Unlikely... I think that even if that would happen it would look different. What we got here is a mess. However, I can't think of a different explanation right now o_O.

After some further analysis with umage we came to a conclusion what possibly happened. The results of my first analysis didn't make sense because they were based on assumption that turns 46 and 47 are consecutive turns. However, this assumption is false.

Turns 46 and 47 are both follow-up turns to turn 45. They are a parallel alternatives that both were created by playing Dark templar in turn 45 by SVirt.

How could this happen? As Fithz Hood pointed out we do have a system that prevents redundant requests, so you can't play one card multiple times (by pressing F5 for example). However, it seems that both these requests were somehow processed simultaneously - i.e. the second request started before the first request finished saving the game state.

Weird. Even weirder that he apparently saw the one result and later refreshed to see that it had been replaced by the other.

I'd still like to know if there have been other cases of this. I imagine that most players would never have noticed unless, like in this case, they saw one result and then had it updated to the other.

Schrödinger's bug

Well, it takes couple dozens of milliseconds to update game state. If you can create a second request in that time interval this can happen.