phosit

We now try to release roughly at the same date each year. We don't wait till all issues with that milestone get implemented. We would push them back. Almost certainly there will be no new civ in the next release.

I don't want to argue in favour of removing support for devices, but... Most arguments are silly (or in Lundukes language: hilarious) "There are issues reported for drivers with no known issues." Yes that's why the issues are reported... "The year 2002. Very very recent in my opinion." When the hardware support is now removed and the already released kernels get supported for maybe 5 years (I think that's how that works), It will be supported for 30 years. I wouldn't call that "Very very recent". "The drivers haven't do be updated in years and years and years" When there is no one maintaining it there is software rot. "Wasn't the point of using AI to make development more efficient‽" AI is used to find bugs and to write code. When the kernel developers wouldn't use AI to find bugs someone else would. When they wouldn't use AI do write code - yes - they would be less efficient. "Back in the days every week we said that something became supported by Linux" Maybe you could do the same thing now.

It would also be nice when they are weaker/capturable from that side.

The starting position and the lakes are similar to "Pyrenean Sierra". Maybe you can base your map on that.

I don't understand the argument. The difficulty of AIs is independent of teams.

What's the point of more teams? When there are more then 4 teams at least one player is a lone in a team. So it has the same effect as to assign no team to that player.

I think that the current Egypt(3v3) should be turned, so that the fertile land is in the foreground and the mountains in the back.

Nice that there are already two starting positions for Petra

Removing an invocation of `pickRandom` doesn't invalidate another. (Worst case is that in DE the AI might not be serializeable.) I tried to bisect it but wasn't successful there are always like 10 units idle.

Sorry for the late reply. Maybe you can find something in this PR. In there are many commits, so bisect should help a lot.

I can't reproduce that. But I noticed that the map isn't revealed after loading.

That's indeed an issue. Thanks for reporting.

That's because it first chooses the placement, without knowing whether the map supports it.

To fix the issue, the map has to somehow expose the restrictions. Those restrictions could then also be read by the random selection algorithm. At least that's how I think of it.

Oh, so you choose random as placement - I understand now. I think it would be a hard to delay the random selection to the mapgen script. One would also have to change replay-summary: there it shouldn't be displayed as "random". The more complete solution is the one stated in the ticket.