Only start Quartz when there are some @Scheduled annotations

[SlyngDK]

…en there are some Schedules.

Fixes: #4243

[mkouba]

Looks good.

[gsmet]

I rebased and force-pushed an adjusted message to try to get that one into 0.25.0.

[jaikiran]

Unless we provide a way to schedule a job programatically...

@mkouba, Looking at the code, it looks like the io.quarkus.scheduler.Scheduler can be injected and can be used programatically to schedule jobs https://github.com/quarkusio/quarkus/blob/master/extensions/scheduler/runtime/src/main/java/io/quarkus/scheduler/Scheduler.java.

The implementation of these APIs, require the scheduler to be started before calling something like startTimer. So if we don't start scheduler, in the absence of a @Schedule, I think that would mean the scheduler can no longer be used programatically.

[gsmet]

CI failures are unrelated, merging.

[gsmet]

Thanks @SlyngDK !

[mkouba]

@jaikiran I've completely forgotten about this one! In theory, we could detect Scheduler injection points too. In which case the only problematic scenario would be "programmatic lookup of a Scheduler + programmatic scheduling" combo.

[jaikiran]

In theory, we could detect Scheduler injection points too.

Yes, that's a good point.

In which case the only problematic scenario would be "programmatic lookup of a Scheduler + programmatic scheduling" combo.

So I had a look at what prompted this change and it appears to be this enhancement suggestion (a valid one IMO) - #4243. I think if that has to be solved and yet be able to start the scheduler eagerly, perhaps a better way would be to implement a (lazy?) org.quartz.spi.ThreadPool which doesn't start those threads eagerly.

[mkouba]

@jaikiran Hm, maybe we should reuse the internal thread pool. Pls file a new tracking issue.

[jaikiran]

@mkouba Done - #4606