Incompatible battle replays across OSes

[alexti]

I've seen this problem many times in MP game I play now and I'd like you to consider the possibility that it's not just the replay being incompatible between the platforms.

In one battle I had won with only 2 mercs left (of some weak variety). Usually these mercs rout way before they all get killed, so I can't realy imagine a scenario in which it would it be possible to win with only 2 of them remaining (opposition consisted of some indy forces). Of course, in Dominions it's impossible to foresee everything and there may be some explanation I'm missing. Replay was showing my mercs retreating pretty quickly after losing few comrades. And I had few other battles with weird results. Everytime, development in the replay looks quite logical and probable, but battle report sometimes produce results which I couldn't interprete at all (meaning that I couldn't imagine how the battle would have to be going to produce such results - as in that example with 2 mercs left).

From my cross-platform programming experience I would have considered that maybe some variable is left uninitialized and on different OS/compilers it gets initiaized differently which produces consistently wrong results in the case of one OS (or different wrong results in different OS). This could explain strange results due to altered battle mechanics.

[SurvivalistMerc]

I actually have a scenario in which I could have won that. My serpent dancers all radiate powerful fear, and I had a good number of them in the battle. You by contrast had no priests. I had those cute little poison slingers. And my sauromancers were casting terror.

If your longbowmen were terrorized.... All I mean to say is that the outcome we got was possible. Not likely but possible. I lost a number of high defense serpent dancers there by the way. Not that they are equal to your giants.

[alexti]

Another wild thought here. If you recompile Dom2 with different byte alignment and try to compare if the battle replay goes the same way in the old executable and the new one (on the same machine). If they're not going identically, surely there must be some uninitialized variables or memory overruns.

/Alex

[Johan K]

I have just installed and tried 'valgrind'. Seems like an excellent program to find variable that are used unitialized. Let's just hope it finds something interesting.

[BugRoger]

We also experienced inaccurate battle replays. In fact the battle replays have absolutely nothing to do with the battle report.

We're playing a 2-player MP game with a Linux host and 2 Windows clients.

I hope you can find that bug. It's not really *that* important but the battle replay is really helpful for tweaking the army setup/spells. Especially for us newbies...

[PDF]

Methinks that the totally inconsistent replays observed by PhilD rather show some problem/bug with the replay module itself than some n-th precision difference in the combat algorithm between various C++ compilers...

[BugRoger]

We experimented a little bit more and don't have problems when hosting and playing on windows. I'm almost certain that this is a cross platform issue.

[tinkthank]

Originally posted by PDF:
Methinks that the totally inconsistent replays observed by PhilD rather show some problem/bug with the replay module itself than some n-th precision difference in the combat algorithm between various C++ compilers...

How do you people understand this stuff?

[Norfleet]

It's technomancy. If you don't understand what any of it means, you're probably not going to unless you take up computer programming. Try researching the Technomancy school. This should make sense to you once you reach level 6 Technomancy or so.

Also, didn't this thread die ages ago? I recall the causes of the battle replay incompatibility was caused by differences in order of execution in the MR if() check?

[ March 18, 2004, 07:40: Message edited by: Norfleet ]

[tinkthank]

Ja sorry I was searching. Shouldnt have replied. Sorry.