From 3fd1150e282f28c18fa685b17960aaf04c20def2 Mon Sep 17 00:00:00 2001 From: dffdff2423 Date: Thu, 17 Oct 2024 18:46:50 -0400 Subject: [PATCH] add forking guide --- src/SUMMARY.md | 1 + src/en/general-development/tips/forking.md | 94 ++++++++++++++++++++++ 2 files changed, 95 insertions(+) create mode 100644 src/en/general-development/tips/forking.md diff --git a/src/SUMMARY.md b/src/SUMMARY.md index 8ac26f89c..facf4026f 100644 --- a/src/SUMMARY.md +++ b/src/SUMMARY.md @@ -35,6 +35,7 @@ General Development - [Writing Guidebook Entries](en/general-development/tips/writing-guidebook-entries.md) - [Config File Reference](en/general-development/tips/config-file-reference.md) - [YAML Crash Course](en/general-development/tips/yaml-crash-course.md) + - [Forking Tips](en/general-development/tips/forking.md) - [Feature Proposals](en/general-development/feature-proposals.md) - [Feature Proposal Template](en/templates/proposal.md) - [Expected Team Decorum & Usage](en/general-development/feature-proposals/expected-feature-proposal-decorum.md) diff --git a/src/en/general-development/tips/forking.md b/src/en/general-development/tips/forking.md new file mode 100644 index 000000000..3e11defeb --- /dev/null +++ b/src/en/general-development/tips/forking.md @@ -0,0 +1,94 @@ +# Forking Tips + +A general set of forking tips and agreed upon best practices from fork devs. If you are a fork dev that would like to add something here please open a PR. + +It is strongly recommended that you have some experience with SS14 and RT before creating your own fork. + +This document only describes the technical considerations of forking. Keep in mind that any good fork will require an administration team and policies. + +## Creating Your Repo + +You should not create your project's reposistory with the github fork button. This is because github only permits one fork of a repo per account and forks of your repo will be counted the same as forks of your upstream. + +Instead, download your upstream's git tree locally on your computer and create a blank git repo in the organization for your project. Then simply push your changes to the repo. Anybody who wants to contribute will need to fork your new repository instead of your upstream's. + +## Organization +### Put your code in dedicated server folders. +A highly recommended pattern is to put your server's code (alongside an optional but recommended LICENSE file containing your code's software license) in its own top level folder within the major sections of the codebase. + +For example, all of the `Content.*` folders would have a `_ServerNameHere` folder inside, in which all of your server specific code would go. Avoid mixing your own code with upstream's in that regard to avoid potential merge conflicts down the line. + +### Namespace your components (and any other serializeable types) +All serializeable types are in the same global namespace. To prevent conflicts you should prefix them with a short identifier of your fork. For instance, if your server was called "Foo Bar" you might prefix all components with "FB". Keeping it short is advisable because you will be typing it a ton. + +### Namespace your prototypes + +In a similar vein to namespacing serializable types, prototypes also share a singular global namespace. Strongly consider giving all prototype names a prefix to avoid conflicts. For example one might give an entity the ID `FooMyEntity` instead of `MyEntity` if the shorthand of your server was "foo". + +### Merge conflicts aren't that scary. +In short, when modifying existing code, conflict avoidance techniques are generally *bad* practice, as they often allow your code to continue compiling at the cost of it not actually working correctly anymore. In general if you get a merge conflict, it is likely not for no reason and you should review it manually instead of trying to incorporate avoidance techniques. + +```admonish info +While this is mostly true, avoiding conflicts in, say, lists of elements, is generally a good idea (as they're going to conflict unnecessarily more often than not.) +This is usually as simple as just putting your entries at the start of the list, or in a separate part of the list separated by whitespace. +``` + +When modifying upstream code it is recommended to leave a short comment explain *why* you changed what you did in order to make upstream merges easier. An example of this would be `// FOOFORK: Changed Bar to 2 instead of 1 becase it breaks everything if it is 1.` + +### Avoiding issues in shared enums. +A big example of this is the AdminLog enum containing all the kinds of admin logs. Pick a recognizable decimal (or hexadecimal) prefix, positive or negative, for your values, and stick to that. Ideally, pick one randomly, so you don't have issues merging other fork's changes if you want to cherry pick. + +## CVars +### Namespace your CVars! +CVars can be as many tables deep as you want, for example `foo.respawn.time` is a valid CVar key and would map to the following: +```toml +[foo.respawn] +time = 360 +``` + +This helps disambiguate where a CVar came from. Additionally, consider migrating any CVars you make significant changes to the behavior of (say, adding new UI layouts or changing the enum keys) to your prefix. + +### Use your own `CVarDefs` instead of CCVars +The `CCVars` master class is something of an antipattern even upstream. You can declare new CVarDef classes anywhere you want, and put your CVars there instead of in CCVars. +For example: +```csharp +[CVarDefs] +public static class MySubsystemCVars +{ + /// + /// This is my CVar! + /// + public readonly static CVarDef MyCVar = CVarDef.Create("foofork.subtable.mycvar", false); +} +``` + +It may be advisable to have a CVarDefs per major subsystem, for example you'd put your world generation cvars in the same folder as the worldgen code itself. The name of the static class does not matter. + +You **can** put a CVarDefs class in only the client or only the server, but be aware that this will result in the CVar being logged as missing if it winds up in a config file loaded by both sides. + +### **DO** edit the defaults in code! +It's likely tempting to modify or introduce a config preset to modify CCVars defaults without editing the class itself. This kind of falls in line with [Merge conflicts aren't that scary.](#merge-conflicts-arent-that-scary), doing so leaves you open to upstream changing the default in a meaningful way (for example, changing what a number actually represents, like minutes to seconds or vice versa.) and getting stuck with a bug on production if it goes unnoticed. + +The conflicts are trying to tell you something, if they happen. + +### But don't make the defaults server-specific. +CCVars is not your server's config file, configure your server on your server in `server_config.toml`, not in the codebase. Change defaults if the default breaks the game, enables mechanics you want disabled, or breaks your design, not to set your discord link. + +## Solutions & Assemblies +### Extra DLLs are not mods. +RobustToolbox features a "mod-loader", allowing you to separate functionality into additional DLLs. This is how some third-party libraries work. However, this functionality deliberately doesn't include support for resources, and shouldn't be used like some sort of drag-drop modding toolkit unless you want the game to be about as coherent as gmod. However, use it for libraries, through submodules, and it'll prove useful. + +### Make sure to change Content.Packaging to include new assemblies. + +If using a third party library, make sure to add the name of its DLL(s) to ClientPackaging.cs and ServerPackaging.cs as appropriate, to ensure release builds of the game include the necessary files. + +### Assemblies must follow the `Content.*` naming scheme. +In short, if you don't do this you will be met with sandbox errors, and your game will not function. + +## Database Layout + +EFCore is not really made to be used in the way that you will be using it as a fork developer, if you can avoid making DB changes you should. To avoid a number of bugs (most notably [this one](https://github.com/dotnet/efcore/issues/24834)) it is strongly recommended to not modify your upstream's tables. Instead, you should make a corresponding table for your fork's data that is in a one to one relationship with the original table. + +Any migrations you add should be namespaced in the same way as prototypes and components. While the likelihood of a migration collision is basically nonexistent, it makes it clear which migrations were added by your fork. + +Finally, make sure that you test your changes and migrations on both postgres and sqlite to avoid data loss. Remember to regularly take backups of your database.