|
Post by zaimoni on Dec 8, 2019 18:20:16 GMT
For those of you tracking the GitHub (and thus seeing the glacial commit rate): the next user-visible planned feature is I-435 (the highway encircling the city). This is soft-blocked by CPU/turn (larger map will cause more extensive pathing costs), and the game is in a position where all known profile-guided optimization targets are either numerically negligible a priori, or cause measurable slowdown. If I have to commit a save-breaking change, there are other things I could land.
|
|
|
Post by zaimoni on Nov 14, 2019 5:29:34 GMT
There doesn't seem to be any non-ideological hosting for Mercurial repositories, now that BitBucket is going away for that purpose. (That said, I'll still be using Mercurial for what it does cleanly.) There was no way to rationally decide between GitLab and GitHub.
|
|
|
Post by zaimoni on Sept 29, 2019 15:27:05 GMT
|
|
|
Post by zaimoni on Sept 29, 2019 5:11:51 GMT
Other notes re .NET Core 3.0: - System.Diagnostics.Activity is new (and can collide with djack.RogueSurvivor.Data.Activity). Mechanical workaround.
- Unlike Windows Presentation Framework, System.Random is not serializable in .NET Core 3.0; this is a blocker for porting to .NET Core.
|
|
|
Post by zaimoni on Sept 28, 2019 17:37:55 GMT
|
|
|
Post by zaimoni on Sept 27, 2019 21:39:47 GMT
This change isn't actually landed in RS Revived's trunk (caught in testing and locked down). It would have been useful to switch Actor over to ISerializable (this enables cache variables that don't reach the savefile, which would be useful for dealing with some profile-hot code called from pathfinding). But ... no. Map::m_ActorsList suddenly breaks down (correct length, but all entries null -- crash during load). No relevant search engine hits (there are a few from 2012 or so but they are not from "doing things by the book"). Diagnosis steps remaining: - Does force-upgrade of Visual Studio fix this? (no)
- If force-upgrade failed, does full rebuild fix this? (no)
|
|
|
Post by zaimoni on Sept 12, 2019 21:02:05 GMT
The AI augmentations needed to get reasonable behavior out of the hospital are extreme. (Ideally, the doctors and nurses would be a civilian sub-faction that try to use the hospital as their base.) You have to be able to chain sub-objectives and actually estimate how many turns a subobjective would cost; not even RS Revived has a prototype of a behavior using this functionality (at this time).
In general, CivilianAI needs to have the ability to path to squares not seen recently to avoid being resource-starved when starting in the hospital or wandering into basements/subways. Preferably in a way that isn't a CPU and RAM hog.
So, no it's not you. BehaviorExplore and BehaviorWander just aren't enough to get the patients out of the hospital (the zone buffer isn't nearly large enough).
|
|
|
Post by zaimoni on Aug 27, 2019 10:45:49 GMT
Yes. That's why turns are at least three times faster Day 14+ than Day 0 Hour 2. (I need to work out a testing procedure that monitors turn times at various stages in the game).
|
|
|
Post by zaimoni on Aug 26, 2019 13:39:15 GMT
All versions past 0.9.9.5 are known to run slow. It's a side effect of the police actually doing their job.
|
|
|
Post by zaimoni on Aug 22, 2019 12:14:27 GMT
Also, why do I starve to death in a day when it should be something like 3-4 days in real life? Ahem...we're not tracking thirst explicitly. That should be lethal in ~1 day. In-game food isn't just calories, it's also the water not being explicitly modeled. Our "main competitor" had a horrible time redoing their hunger mechanic back in June 2019; they didn't take into account that two-thirds of the general U.S. population never experiences proper hunger, so there was no way a medically realistic implementation would "feel right". That forty days or so to critical starvation you read about is for being intentionally sedentary; that's not relevant here.
|
|
|
Post by zaimoni on Aug 22, 2019 11:50:24 GMT
I'm toying with the idea of replacing the current unique actors (Duckman, JM, Santa, etc) with a new unique set that better fits the theme. I appreciate the feedback regarding their possible invincibility; I'll look into that for a quick fix in the meantime. Did you import the RS Alpha 10 invincibility until seen feature? (It did not fit with the way I want to take RS Revived, so was one of the won't-migrate features). That sounds like the sighting that cancels invincibility doesn't get processed.
|
|
|
Post by zaimoni on Aug 21, 2019 19:27:21 GMT
The Feral dog content in RS Alpha 9 was decidedly incomplete. I don't have a clear visual on how they were supposed to work.
|
|
|
Post by zaimoni on Jul 28, 2019 4:15:03 GMT
For those checking trunk: the recent gymnastics have been trying to line up changing the Point type the main game uses "cleanly" (with a minimum of manual casting to short i.e. 16-bit integers). The problem is save game load time: 120MB was Day 16 hour 14; 138MB was Day 14 hour 3. I don't expect this to be a major issue for Staying Alive or Rogue Survivor Alpha 10.1; the police ai support is expected to be the cause of the huge save files for RS Revived.
Right now I don't have a hard block on introducing the highway encircling the city (the city is large enough to make this a reasonable expectation, and I want this in before attempting helicopters), and there has been a map-generation break landed already.
|
|
|
Post by zaimoni on Jul 5, 2019 0:57:49 GMT
Just cut the July 4 2019 unstable release. - new text label SIMULATING when it is not your turn. Not 100% reliable; if things look weird try the R)un command to force a screen redraw.
- same screen space also has the SAVING label when saving a game. This replaces wiping the displayed messages.
- Counted walk command now available; 1) through 9) steps.
- Considerably fewer cases of living ais inexplicably stepping between the same two locations.
|
|
|
Post by zaimoni on Jun 8, 2019 0:52:48 GMT
Just cut the June 7 2019 unstable release. Burning cars now incinerate air drops (and thus do not elicit truly ugly pathing issues).
I'm not sure how "stable" pathing is, but the crash-on-period-2-move-loop only happens in debug builds anyway.
|
|