Let's make VerboseProcess closeable about jcabi-log

[yegor256]

I don't get, what exactly is a problem and how to reproduce it?

[longtimeago]

@yegor256 Well, I know how it looks in thindeck project.

1. Switch thindeck project on jcabi-log 0.15
2. Run the build
   You will see a lot of such exceptions in traces (build would still pass).

How to reproduce it in jcabi-log project? I don't know.

I just try to be proactive and don't wait for the problem would be seen in thindeck/other_proj master

[yegor256]

if it's not reproducible here, how do you know the problem is related to jcabi-log? we should find a way to reproduce it here, then we can fix...

[longtimeago]

@yegor256 You are right

[longtimeago]

@yegor256 So, are we going to take care of the issue? Should I close it? Is there anything I can do here as a reporter?

[yegor256]

it's up to you. if you can find a way to reproduce it, please post it here, we'll turn into a real "bug" and someone will fix it. if you can't find a way, just close it.

[longtimeago]

ok, I'll try to reproduce it

[longtimeago]

@yegor256 I think I understand what is the problem.
We observe the exception because we are trying to read from stream but it's closed. Why is it closed? Because process was destroyed!
Take a look at NginxTest.createsHostsConfiguration https://github.com/yegor256/thindeck/blob/master/src/test/java/com/thindeck/steps/NginxTest.java#L137
We call sshd.stop(); which is https://github.com/jcabi/jcabi-ssh/blob/master/src/main/java/com/jcabi/ssh/SSHD.java#L146 this.process.destroy();
So, process is destroyed, but decorator VerboseProcessknows nothing about that! That's why it's internal monitors fail to read from closed stream.

How to fix that?
There are two points of fix:

1. [jcabi-log] VerboseProcess should be responsible for closing 2 internal monitors and then closing decorated process. IMHO, good idea if VerboseProcess would implement Closable
2. [jcabi-ssh] Make SSHD.close() to call VerboseProcess.close instead of Process.destroy

What do you think?

[yegor256]

sounds interesting indeed. let's change the description of the ticket and make VerboseProcess closeable

[dmarkov]

I'll find a developer for the task soon...

[dmarkov]

thanks for this report, I added 15 mins to your account, in transaction 46857728

[longtimeago]

@yegor256 Can someone take care of the issue?

[yegor256]

@dmarkov please assign @carlosmiranda to this task

[dmarkov]

@dmarkov please assign @carlosmiranda to this task

@yegor256 OK @carlosmiranda please proceed, this task is yours

[carlosmiranda]

@longtimeago I'm not sure I follow how this will resolve the problem. The monitors are daemon threads and not executors, so they're not "closeable" either. However they will throw an exception, if the process is destroyed since the streams will be closed. It's also possible for the process to end without destroy() being called on it.

That said, I don't know of a way to determining "elegantly" whether the process itself has terminated. We can call process.exitValue() and see if it throws an exception, but that's an ugly way to do it. :(

What do you think?

[longtimeago]

@carlosmiranda Let's get the light.
Main goal - make VerboseProcess closable. When VerboseProcess client call VerboseProcess.close() VerboseProcess should gracefully stop it's monitors and THEN call destroy() on wrapped process. So far so good?

You are right, monitors are daemon threads. To gracefully stop them we need to create a Thread which can be stopped from outside (see http://www.java2s.com/Code/Java/Threads/Thesafewaytostopathread.htm for example).
We would also have to store those monitor threads as fields in VerboseProcess.

[carlosmiranda]

Pull request submitted at #42.

[carlosmiranda]

@longtimeago it's now closable in #42. Issue with monitor threads will be continued in a puzzle. Can we close this now?

[longtimeago]

@carlosmiranda sure, thank you very much!

[dmarkov]

@carlosmiranda 30 mins was added to your account, many thanks for your contribution!

[dmarkov]

once 38-9367cb1e puzzle is resolved, this ticket will be complete

[dmarkov]

the last puzzle 38-9367cb1e/#46 solved