-
Notifications
You must be signed in to change notification settings - Fork 420
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Implement changes to client auto-test threading #4256
Implement changes to client auto-test threading #4256
Conversation
...ase/src/testmodClient/java/net/fabricmc/fabric/test/base/client/FabricApiAutoTestClient.java
Outdated
Show resolved
Hide resolved
…ts, reintroduce MinecraftClient parameter
.../src/testmodClient/java/net/fabricmc/fabric/test/base/client/mixin/MinecraftClientMixin.java
Outdated
Show resolved
Hide resolved
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This looks really solid overall! 👍 I won't pretend to understand all of the multithreading logic, but we can see from the tests that it works and I can also see that you've put a lot of effort into making sure every edge case has safety checks included.
public static <T, E extends Throwable> T computeOnClient(FailableFunction<MinecraftClient, T, E> action) throws E { | ||
MutableObject<T> result = new MutableObject<>(); | ||
runOnClient(minecraft -> result.setValue(action.apply(minecraft))); | ||
return result.getValue(); | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Suggestion: Calling it getFromClient()
instead of computeOnClient()
would make it clearer that this method returns a value while runOnClient()
doesn't.
Your use of compute
here isn't wrong, I believe the Java API does something similar, but I think many people use run
and compute
synonymously and wouldn't immediately grasp that this is the difference between them. get
vs run
makes it really obvious.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm pretty sure this is the more common terminology, although I'm not super attached to it. But it should be immediately obvious what the difference between the two functions are from the parameter type
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I do like simple language, but yeah compute
is pretty common too. I don't think parameter/return types are usually the first thing I would look at when working with a new API, but then again this will eventually have Javadoc that can make it obvious more quickly.
.../src/testmodClient/java/net/fabricmc/fabric/test/base/client/mixin/MinecraftClientMixin.java
Outdated
Show resolved
Hide resolved
private <T> CompletableFuture<T> submit(Function<MinecraftDedicatedServer, T> function) { | ||
return server.submit(() -> function.apply(server)); | ||
} | ||
|
||
private <T> T submitAndWait(Function<MinecraftDedicatedServer, T> function) { | ||
return submit(function).join(); | ||
} | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why are there no runOnServer()
/ computeOnServer()
helpers here? I mean, I can't say that I've used the old TestDedicatedServer.submitAndWait()
in actual test code, but this seems like something that more server-heavy mods would find useful in their tests. Obviously one can run ThreadingImpl.runOnServer()
directly instead, as you've done above, but it feels inconsistent to have these helpers on one side and not the other.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
They were private and unused so I deleted them, the equivalent is in ThreadingImpl
. We can think about what helpers are useful to expose in a future PR
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
OK.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ill be honest I dont full understand the impl, but the principle is sound.
public static Runnable taskToRun = null; | ||
|
||
public static void enterPhase(int phase) { | ||
while ((PHASER.getPhase() & PHASE_MASK) != phase) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is the only bit that is midly confusing to me (I think), is this method blocking until the current phase has been reached from somewhere else? My tottal lack of knowledge of Phaser doesnt help
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This method blocks until the given tick phase has been reached on all threads registered to the Phaser
("parties"), while allowing other parties to advance through the other tick phases (which is what the loop is for). The arriveAndAwaitAdvance
is a barrier that all parties must call before they are all unblocked at once.
A Phaser
maintains a counter called the "phase" which increments each time a barrier is crossed. It starts at 0 and when it surpasses Integer.MAX_VALUE
it resets back to 0 again (negative values happen when the Phaser
is terminated, which is not functionality we're using). Using the lower 2 bits of this counter that already exists tells us which of our 4 tick phases we're currently in. It's possible to track that separately by subclassing the Phaser
and overriding its onAdvance
method, but I figured why not take advantage of a counter that already exists.
Implement changes to client auto-test threading as discussed on Discord. More detailed implementation notes can be found on the docs to
ThreadingImpl
.I didn't put much thought into class naming, or what should or shouldn't be in the API, etc, as this is all still internal, and deciding on a nice clean API is left for future work.