ARTEMIS-3327 Removing unecessary block operation on journal append record

[clebertsuconic]

No description provided.

[clebertsuconic]

this is simply an optimization. There are no semantic changes on this commit.

[jbertram]

My reading here is that the result wasn't actually used for anything so there was no reason to keep it. Is that how you see it?

[clebertsuconic]

@jbertram that's right

[brusdev]

@clebertsuconic was the result used to wait the appendExecutor execution?

[gemmellr]

Depends on the provided values of 'sync' and 'callback'. The called newSyncAndCallbackResult(sync, callback) method creating the future did this:
return (sync && callback == null) ? new SimpleFutureImpl<>() : SimpleFuture.dumb();

If asked to sync and there was no callback, it returns a 'proper future' that could be used to actually wait for the execution, and detect failures from thrown exceptions passed via the fail method.

If not asked to sync, or a callback is given, it returns a 'dumb' future that doesnt wait at all and just returns null immediately, ignores completions/failures etc.

I dont know what the sync and callback values actually are for this situation but I guess from that they were eitehr always false or non-null. Might be worth adding a check to that effect though.

[clebertsuconic]

@gemmellr / @brusdev / @jbertram I was thinking only about the artemis use case here, where we always use this method with sync=false.

I know some people is using the journal directly (who actually requested me to make the journal an independent project a few times).. and they actually will use the sync version.

I will just revert this method. thanks for the comments.

[gemmellr]

Might be worth adding a comment, and test, to avoid a recurrence.

[gemmellr]

Revert and test addition were done via #3647