[401] stop piper thread via flag

[ghentschke]

fixes #401

[jonahgraham]

I can't add line comments because the lines I want to comment on are unmodified:

For:

final Runnable c = cancelation[0];
if ((c != null)) {
	synchronized (c) {
		c.notify();
	}
}

I don't think this is needed anymore as I don't think anything is waiting on Runnable c. Once you do that, I think a bit of simplification is trivally possible as final Runnable[] cancelation is no longer needed and the method can just return stopper.

For:

} catch (Throwable t) {
	throw sneakyThrow(t);
}

I think this needs to check for stop and suppress the exception if stop is requested. This is because at the point that stop is requested the code may be either inside of the read() call or just after you checked for stop but before read is called. In that case you want to suppress the now expected exception. E.g:

} catch (Throwable t) {
	if (!stop.get()) {
		throw sneakyThrow(t);
	}
}

This structure I am recommending here is based on OutputStremMonitor (link to it's catch block)

[jonahgraham]

I like that you added a thread name. Can you mention CDT in it?

Suggested change

	errorStreamPipeStopper = new AsyncStreamPipe().pipeTo("LS stderr pipe", //$NON-NLS-1$
	errorStreamPipeStopper = new AsyncStreamPipe().pipeTo("CDT LS stderr pipe", //$NON-NLS-1$

[ghentschke]

I think I fixed all issues mentioned above

[jonahgraham]

There is a small amount of dead code still in this patchset hidden by the @SuppressWarnings("all") which isn't needed anymore. Specifically output cannot be null where it is checked in the catch block. OutputStremMonitor can have some nulls in the loop because it uses non-final fields instead of final locals.

[jonahgraham]

BTW this last review is of very much diminishing returns, the current code operates correctly, just has some dead code.

I also accidentally pushed a change to your branch and then removed it: