RS revived

[zaimoni]

Note that if you start a new game of Rogue Survivor Revived with the --socrates-daimon option, you will have a cheat map command available (CTRL-I by default). Once the required group spawns, use the cheat map to identify who you want to play and the --PC option to take control of them. Do this in two stages: --PC the leader, save, close out, then --PC all followers if that is the plan. [If you --PC a follower whose leader is NPC, they will no longer be a follower.]

[chuckokokos]

Neat, gonna try it out. Good luck with the project dude.

[zaimoni]

Unstable release cut Jan 16, 2018. The revised documentation is not completely accurate (the actual waypoint saving is not in, but the far-look for setting the waypoints is in).

The primary push was to verify that a change in inventory management priority representation (to: useless, insurance, want, need) did not contradict the legacy code in an unfixable way.

• new self-order: medicate sleep.
• Getting ALT keypresses seen at all was painful (the required documentation is not in a logical location) ALT-I is still too unreliable to use, so others of that class are assumed unusuable. The other ALT combinations should work when explicitly enabled (ALT-SHIFT-I, ALT-CONTROL-I tested)
• becoming tougher is no longer an injurious experience that immediately requires medical aid.

[zaimoni]

Father Time has landed in trunk (some polish remains to land, including his easter egg for New Year's Day), and will be included in the next test game. I'm evaluating how "cinematic" he should be -- or rather his scythe, which is a lame weapon (slightly worse than an iron golf club) that gets martial arts skill bonuses. I may need to PC-ize him for testing purposes and/or provide an option for force-spawning him.

In general, any vaporware B-movie capabilities of the scythe will require some level of martial arts skill to use. B-movie capabilities, starting with but not limited to range 2 melee, may have to be cut as too close to Cataclysm DDA in power level.

Undecided: is Famu Fataru's katana also a martial arts weapon, and if so how much of its current stats require martial arts?

[MP]

Re katana; it does make sense to need MA. Unlike, say, an axe, I understand bladed weapons really take skill to use effectively. Especially when it comes to zombies, who wouldn't be affected much by anything but decapitation.

[zaimoni]

Agreed. I have a very strong emotional bias towards katana as MA weapon, just don't have a good intuition for how it affects game balance. [ I don't think either National Guard or Those Guys are pre-qualified for the martial arts skill even if boot camp (Nat. Guard) or the weeks surrounding Hell Week (Those Guys) have a week or two worth of review.]

Working notes: unskilled katana still has to hit harder than a combat knife, so there isn't that much room to transfer the damage part. I'm pretty sure most of the to hit needs to be transferred to MA on conversion, and that the katana has to do noticeably more damage than the scythe. MA katana would qualify for B-movie capabilities that were similar to those implemented for the scythe.

[ldfzm]

Hey zaimoni - I just wanted to say thanks for keeping the game alive!  I discovered Rogue Survivor a long time ago and had the sudden urge to play it again yesterday.  Since the version I had installed was RS Alpha 4, I decided to see if there was a newer version and came across the post that led me here!  I'll definitely be checking out a much newer version of the game later

[zaimoni]

Once Staying Alive gets re-released, take a look at that as well.

[zaimoni]

"sort of" in reply to a comment in the Staying Alive thread, but off-topic there:

Yes, I've been considering a fourth gameplay mode (instant zombification+corpses). How it should work when a shambler+ bites a corpse isn't clear.

The source code comments about a new level for the CHAR Underground base are highly speculative. Details depends on:
* What makes the invasion ley lines valid? Neither remote materialization/teleporter technology, nor dimensional gate technology, is anything I would expect given the inherited setting.
* Does "skeletonization" from being killed by a skeleton, or having a corpse completely devoured, make sense?
** Minimum # of plausible direct-conversion agents is four. (I am assuming nanotech gone wrong) Two in very limited supply (that deleted the CHAR Underground Base when it escaped quarantine), two weaker ones that are responsible for the main z invasion supply
** It's likely the z reserves staging area needs to be an abstract space for now. Cf. They Are Billions.
** it would be useful to restructure the savefile, so that the total # of ActorModel instances covered was in the savefile.

[zaimoni]

I've had a commit thaw the past few days (not that work schedule is really cooperating, so once I stop crash-dieting the thaw will resume). Focus has been cosmetic issues and doing some nightmare mode testing on 3x3 size 30 vintage mode (not what I want to balance the game for, but it does exercise a lot of code I otherwise don't touch much).

It is now known to be technically possible to save the RNG state without relocating the RNG from the Rules object to the Session object (proof of concept was first making the model created count survive save-load, then restructuring the savefile to save the current player, rather than his map.) I'm unsure if this makes sense: there is a theoretical risk that if the C# lock primitives fail, the RNG would end up in a corrupted state that returns the constant zero. (Yes, in C++ terms C# locking can fail, because double-checked locking is invalid.)

I want to make savescumming less useful (a guaranteed reset of the RNG is very useful for forcing predictable damage, etc.) so I'm likely to proceed with this build option regardless.

[roguedjack]

HEY YOU!! Good job on keeping the game alive!

I'm afraid as others have reported, the game is laggy too on my system (a pretty good win10 machine).

The thing is, not only during the game turns are slow, but even the loading screen and main menu is noticably slower than rs9. I thought it was GDI+ being annoying, as resizing the window improves performance, but RS9 GDI+ and Still Alive GDI+ don't have this problem when maximized, they still run fast.

I downloaded the stable zip and I found that compiling with targeting x86 instead of "any cpu" and release instead of debug helps performance a lot in the loading/main menu screens. I also unchecked "enable visual studio hosting process" in debug settings, not sure if it did anything. The game turns are still slow though. You could try those parameters, it still helps a bit.

RS9 and Still Alive use dotnet 3.5 and I see you use higher version. If you are not suffering from slow turns on your dev computer, maybe there is something funky going on with specific dotnet versions that makes your program runs very slow on some systems. No idea what it could be though (contracts? threads? input reading lag? gdi+ rendering?...).

It is also possible that the sources you obtained from decompiling has annoying junk added by the decompiler, I noticed some methods attributes like "[SecurityPermission(SecurityAction.LinkDemand, UnmanagedCode = true)]" that weren't in the original source.

If you need some help testing fixes, I could have a quick go at some builds to test them and tell you if they make any difference in performance.

[zaimoni]

The ZIPs actually uploaded are: target Any CPU, Release mode. My test games are done with Any CPU, Debug mode, with the artificial message delay enabled and screen maximized. Game load time is indeed awful starting with the introduction of police district sweeping.

The only uploaded release that normally takes longer than 2 seconds per turn on the 7-year-old development system in city size 5x5 district size 50 Classic, is the January 16 unstable; that has a pathfinding correctness fix that spiked it to 3 seconds/turn. Unusual features of the development system: paging file completely disabled, hard drive rather than SSD. The release builds feel twitch-real-time fast on the development machine, when the artificial message delays turned off.

The method attributes were added early on, as compiler guided warning suppression.

Contracts are a very likely suspect for about half second of slowdown, and do need to be removed since Microsoft has openly declared won't-support for MS C# 2017. I'm only removing them as convenient at this time.

[zaimoni]

Yes, mostly purging System.Diagnostics.Contracts fixed the load time problems in trunk. It doesn't do that much for actual play speed on the development machine (placebo effect "feels faster", but can't measure it).

[roguedjack]

Mar 25, 2018 10:19:44 GMT zaimoni said:

Yes, mostly purging System.Diagnostics.Contracts fixed the load time problems in trunk. It doesn't do that much for actual play speed on the development machine (placebo effect "feels faster", but can't measure it).

I can test your builds on my machine if that can help you find the problem, just send me an email or pm with a link to download it. If there is an improvement I should see it right away.
Remember to set build to x86 not any cpu, it really does make a difference on my machine, and I suspect for other users having a slow game too. When i'm talking about "loading screen" being slow, I meant the very first things the game does before going to the main menu like loading images, music, checking directories etc... For some reason this runs very slow when compiled with "any cpu" and compiling for x86 makes it run as fast as rs9 and still alive.

[zaimoni]

Difference between x64 and AnyCPU is untestable on the development system. A sufficient explanation, would be a JIT compiler triggering on the development system but not elsewhere.

What I should do, is change release policy to include both AnyCPU and specific processor binaries.