OOS after rejoin - territory differences on square maps

[temple]

fpre and Palaiologos shared oos logs that were similar to the KunstRaucher ones in #5162 (territory of some trees and grasses was different), but I was able to reproduce theirs with rejointest after recreating the map in atlas. I think the attached is for turn 10.

[temple]

After some investigation with elexis, we found ​Phab:rP21675 responsible. In particular, on square maps (like latium) starting players see territories that ignore passability concerns, while rejoining players see territories that adjust for hills, etc. That's why the rejoining players have the trees in gaia territory rather than player territory.

(However, I haven't yet digested the code to explain why it fails only on square maps. The KunstRaucher replay from #5162 is a similar oos and on the circular mainland map, but we couldn't reproduce it with rejointest.)

[elexis]

Is this a self-correcting OOS and thus primarily occurs within the first 20 turns?

[temple]

No, the cost grid remains all 1's (ignores terrain) throughout the game for the initial players while rejoiners have territories adjusted for terrain.

[temple]

In 21835:

Always call SetPassabilityCircular -- Fix OOS after rejoin on square maps

Differential Revision: ​https://code.wildfiregames.com/D1555
Reviewed by: elexis
Fixes: #5186

[smiley]

21835 was a hotfix which did not fix the bug.

[elexis]

I'd prefer to keep the resolved OOS tickets on this milestone, if it's just for finding it historically.

Please always post the task when opening a ticket, I guess you refer to ​https://code.wildfiregames.com/rP21675#30839 (and set the upcoming milestone only if you know someone who will work on it then).

[smiley]

More or less. Also the fact that if SetPassabilityCircular is called with false as square maps should, the game would still go oos. Would it not? I think I read something wrong.