Investigate io.quarkus.resteasy.reactive.server.test.stream.StreamTestCase#testClientStreaming

[cescoffier]

io.quarkus.resteasy.reactive.server.test.stream.StreamTestCase#testClientStreaming has been disabled in the Vert.x 4 branches due to some random failure on CI.

Also:
io.quarkus.resteasy.reactive.jsonb.deployment.test.sse.SseTestCase#testMultiFromMulti

[quarkus-bot]

/cc @FroMage, @geoand, @stuartwdouglas

[famod]

FWIW, we have seen this test class failing randomly on main as well: #15756

[geoand]

I was periodically able to reproduce this locally but I still don't know what the cause is.

A packet capture seems to show that the server returns the proper data in all cases so judging from that it looks like a client issue.

[FroMage]

Huh, that could be very helpful, narrows it down quite a lot !

[geoand]

I'm looking at MultiInvoker and I am wondering whether the MultiRequest class you added is still needed. Asking because it looks like the bug you mention seems to be fixed in Mutiny.
So perhaps the code can be simplified?

[FroMage]

Probably yes. And this was tested anyway. But we disabled so many tests, it's hard to be sure it's still enabled…

[geoand]

It seems like in the cases of failure,

quarkus/independent-projects/resteasy-reactive/client/runtime/src/main/java/org/jboss/resteasy/reactive/client/impl/MultiInvoker.java

Line 161 in c373306

multiRequest.emitter.complete();

is getting called before

quarkus/independent-projects/resteasy-reactive/client/runtime/src/main/java/org/jboss/resteasy/reactive/client/impl/MultiInvoker.java

Line 170 in c373306

multiRequest.emitter.emit(item);

The thing is that I don't think we even need to have

        connection.closeHandler(v -> {
            multiRequest.emitter.complete();
        });

as

        vertxClientResponse.endHandler(v -> {
            multiRequest.emitter.complete();
        });

will always be called anyway (and it does seem to always be called after the emitting is done).

WDYT @FroMage ?

[geoand]

FWIW, I ran thousands of tests with the above fix in place and they all passed

[geoand]

The same thing exact also applies to SseEventSourceImpl. I will open a PR

[geoand]

Here it is: #16438