Ah, so I think I misunderstood the feature. When I set a value for "Abort the
build if it's stuck", this same timeout applies separately to executor
starvation and the actual build execution time?
I guess what's weird is that we have different jobs that take wildly varying
lengths of time. So I have timeouts for various jobs at 5 minutes, 30 minutes,
1 hour and 3 hours. For a couple of hours during the day for us, it's very easy
to get a 15-minute queue depth. The five minute jobs will then be reported as
stuck while sitting in the queue, but the other jobs won't.
I think that just a better indication of what is really happening is an ok solution.
The intention of marking such jobs in the queue is to let you know that these
are starving for additional executors — it's time to add more slaves.
Perhaps we should change the wording a bit? Like from "stuck" to "starving
executors" or something like that? With perhaps a hyperlink to a page in Wiki
that talks about what you are supposed to do when you see it?
Or maybe we should change the color from red to something less scary, like yellow?