RS revived

[zaimoni]

Unstable release cut; the major fix here is subway generation (it could infinite-loop before game start).

Speed increases compared to the last unstable release are underwhelming.

[zaimoni]

Unstable release cut; savefile break followed by pushing the test game out to Day 0 hour 6 flushed out a number of highly unstable issues, so the prior one was withdrawn.

[zaimoni]

Unstable release cut. It appears that the CPU blowup during the police assault on the CHAR base has been prevented (CPU/turn did not go up materially).

Note the new infix -core ; this one is built against .NET Core 3.1 rather than .NET 4.7.1. Eventually, I'll switch over to .NET Core as the backing runtime as there are a couple of technical changes that require the bytecode extensions used there. (That is, the required syntax won't build for .NET 4.7.1, period.)

[zaimoni]

There's going to be a commit stall for an unspecified time (hopefully only a few days): need to pre-calculate a change plan for savefile-breaking changes. At least one of the following will be scheduled:

• CPU/turn is somewhere below 15 seconds/turn through Day 1 hour 12 consistently, for the reference test game. That's not "low enough" to comfortably introduce the encircling highway, but I have no clear way forward to reduce CPU/turn further while maintaining full cross-district pathing. This highway is "borderline" (final version will have gas stations [pulled from Staying Alive], but it still is in the pre-apocalypse broadcast power zone)
• The member function changes needed to support binding container inventories, to containers, has landed. The actual binding hasn't been committed yet. (I.e., pushing that store shelf doesn't leave behind what's on it.)
• I'd like items on tables and chairs to be "reachable" when adjacent, rather than having to jump on them to access. (This is a radical player UI change; not so radical for ai. The bump move heuristics check for jumpable object before container object.)
• I'd like to route worn armor to its own physical slot, rather than eating a slot in the hammerspace inventory. (Implications: armor remains on the corpse, and thus can be accidentally worn by unevolved zombified humans. It is extremely unlikely to survive the upgrade.)

[zaimoni]

Commit stall thawed (finally). Shaking down container inventories has been messy (and stress-tested the lambda-function pathfinder)

[zaimoni]

Release cycle plans:

.NET 5.0 is scheduled for "stable release" Nov. 2020. This is when Microsoft abandons the standalone Windows Presentation Framework (WPF) (latest version 4.8 series, RS Revived currently on 4.6.1) and will maintain only the one that ships with with the .NET 5.0 Windows sub-target. (i.e., it is a platform shift to re-target the event handler system from .NET 5.0 Windows to .NET 5.0 cross-platform.) The release candidate out currently (RC1) has public relations indicating there is not expected to be an extra release candidate.

While there do not appear to be any killer-app features in C# syntax (jump would be from 7.2 to 9.0), there is a very major change in build procedure available: Ahead-of-time compilation, which almost allows paying the JIT overhead in advance, much like traditional C++ compiling. It also is represented as allowing shipping an *.exe that doesn't require the target system to have .NET runtime installed

So, no later than the Nov 2020 release, intent is:

• Emergency unstable checkpoint for RS Revived (built against WPF 4.6.1)
• Cut over the .NET Core 3.1 build of RS Revived to .NET 5.0 Windows. Determine whether save-load cycling and new game are broken; if broken, fix.
• Adjust build system to only generate the .NET Core 5.0 build.

If I can get both acceptable AI and not-that-awful CPU/turn then I'll start the above against .NET 5.0 RC 1. I don't think parity with RS Alpha 9 (3-4 seconds/turn at Day 0 Hour 0 at maximum simulation detail, etc. for 5x5 50x50) is tenable before cutover to .NET 5.0 Windows (there are just too many low-level optimizations the higher version is represented as having, that the lower versions don't have).

[zaimoni]

Unstable release cut. The official release date for .NET 5.0 is "tomorrow", November 10, 2020. Wanted to make sure there was a binary build before attempting upgrade.

Changes of note since the last checkpoint release:

• Friends are now enemy detectors. You must be able to see them; detection isn't enough. Icons are known-broken (zombies get the zombie detector icon; everyone else gets the blackops icon). The detectee only has to be targetable with the best ranged weapon. It'll trigger even if that isn't in use. You cannot read off an enemy at range five, from someone who only has a shotgun (typically range 3). This does work for melee weapons (with targeting range 1, except for Father Time's Scythe which is range 2.)
• Containers now have real inventory: pushing that store shelf takes the item on it, with it.
• The police are now capable of finding the secret underground base on their own. If you do not get there in time, they will first power it up, and then loot it.

CPU/turn on the development system ranges from 3 to 15 seconds (it spikes up dramatically on the noon refugee arrival, and slightly on the midnight invasion).

[zaimoni]

Some development policy updates:

• New pan-project devlog at zaimoni.medium.com/ . This is a paid membership remote-hosted weblog site (i.e., unlikely to be blacklisted by Reddit for being ad-supported). Plan is to upload commentary about the reference test game, as a walkthrough.
• Interim release policy has been changed: I am now uploading non-release latest-trunk builds to the same release as the last release. (This is the same policy I use for Cataclysm:Z.) These builds are .NET 5.0-windows only. I was hoping Ahead-of-Time compilation would be present, but it turns that was too complex for Windows.
• I have started prototyping a re-implementation of System.Runtime.Serialization. This is now not just improperly advised against, it rates its own compiler warning as of .NET 5.0. Given other questionable policy decisions, I cannot analytically rule out this being removed at .NET 6.0 with no viable supported replacement. Newtonsoft's BSON does not have the required features, at this time.

[zaimoni]

Waiting on inquiry at Reddit to complete regarding UI migration for building on Linux. Summary at github.com/zaimoni/RSRevived/issues/15 . It appears there are zero Microsoft-supported options.

MP , do you have any experience that would bias between the following:

• Avalonia avaloniaui.net/ (also requires Xorg)
• QMLnet github.com/qmlnet/qmlnet (also requires QT)
• Silk.NET github.com/dotnet/Silk.NET
• Uno platform.uno/

[MP]

To be honest I've yet to give it any thought. Though I transitioned my two personal machines from Windows to Linux (Manjaro KDE and Debian XFCE) at the start of March 2020, I haven't put any time towards RS since then.
My vague plan was to take stenodyon's work as a starting point - namely, de-couple RS from MDX, and utilise only GDI to begin with. I definitely hadn't thought beyond that.

Each of the options that you've listed seems promising, but I still consider myself very much a Linux newbie, so I really can't add any value to the decision at this point. To be perfectly honest, I don't expect there to be any appetite for a Linux-based release of RS. Stenodyon put out a working Linux fork, and apart from the three of us it doesn't seem to have got any attention. I would love to see a fleshed out RS on Linux, but I'm heavily biased - it's my chosen OS these days, so I can't see myself putting work into anything but that (ie. no more work on the Windows build).

Long story short, I'm not familiar enough with the Linux landscape yet to offer any valuable guidance. It's a tricky time too, as most distributions have views towards retiring xorg in favour of Wayland in the not-too-distant future. Add to that the fact that there is heavy fragmentation between QT and GTK amongst the major desktop environments. Therefore, if you are indeed keen on taking a shot at a Linux build, all I can suggest is to pick whichever API seems most appealing to you. If nothing else, I'm happy to playtest whatever you might come up with. And if the future lends itself I might get back to chipping in with some code somewhere down the line.

[SaberVS7]

Holy crap, color me shocked to discover there's still people around here. Just swung back after seeing AmbiguousAmphibian's Cataclysm DDA video and realized "Huh, this looks a lot like Rogue Survivor", and then found out about this. Honestly color me more surprised that this of all places is where it's being posted about.

[zaimoni]

Currently thawed on on commits (working on a city-encircling highway for RS Revived as that blocks so many other features). There's also some technical issues that I would much rather not deal with, but feel obligated to due to Microsoft policies.

This, of course, enlarges the reality bubble even more. I will have to disable this build option for release checkpoints if I can't get the CPU/turn down to something "reasonable". (That is, that 5x5 50x50 reference game is actually 7x7 50x50 with the highway included. As usual, RS Revived sees no measurable performance improvements from upgrading from .NET 5 to .NET 6.)

[zaimoni]

Trunk is currently not-crashing I got the .NET 5.0 conversion wrong -- do not copy-paste the example that throws on cancellation, just return. You won't see the player-dead screen otherwise.

CPU/turn is holding at ~10 seconds/turn worst-case (test game ended at Day 16 Hour 21). This is with the encircling highway and the re-positioned refugee entry methods, but no changes to other living event types.

I need to make some judgment calls regarding how much UI cleanup is to be done; release needs to be cut before trying anything radical, like gas stations from Staying Alive.

[burbank]

*offers quiet encouragement to zaimoni from the corner*

[zaimoni]

I cut a checkpoint a few days ago, to lock in some crash fixes before starting on savefile-breaking changes (to try to give the C# JIT a leg up). I expect there to be a drop-dead release when .NET 7/C# 11 is released.

I don't want to commit to what the next iteration will be pushing for, at this point -- depends on what turns out to be "easy from where the game is at". New content (either from Staying Alive, or specific to the encircling highway zoning) will be after fundamental design. (The data design for even motorcycles proved to be awful. If they land on this iteration, they will be PC-only because the AI simply will not be able to cope with how radically different the movement options are.)