Path: csiph.com!usenet.pasdenom.info!aioe.org!eternal-september.org!feeder.eternal-september.org!mx04.eternal-september.org!.POSTED!not-for-mail
From: Knute Johnson <nospam@knutejohnson.com>
Newsgroups: comp.lang.java.programmer
Subject: Re: Strange Socket problem
Date: Fri, 02 Mar 2012 17:18:04 -0800
Organization: A noiseless patient Spider
Lines: 54
Message-ID: <jirrgc$d64$1@dont-email.me>
References: <jiojsa$5ol$1@dont-email.me> <4f512b27$0$26687$65785112@news.neostrada.pl> <jirjpg$rkb$1@news.albasani.net> <jirlma$bfm$1@dont-email.me> <jirmv1$1gj$1@news.albasani.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Injection-Date: Sat, 3 Mar 2012 01:18:04 +0000 (UTC)
Injection-Info: mx04.eternal-september.org; posting-host="mz/LDSJwiWnk3Jnnqg7x+Q"; logging-data="13508"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX18tr14GFsjr/1i5pysoCEbP"
User-Agent: Mozilla/5.0 (Windows NT 5.1; rv:10.0.2) Gecko/20120216 Thunderbird/10.0.2
In-Reply-To: <jirmv1$1gj$1@news.albasani.net>
Cancel-Lock: sha1:PZf8ogxRHw84O5pWj9P8osdxjtM=
Xref: csiph.com comp.lang.java.programmer:12611

On 3/2/2012 4:00 PM, Lew wrote:
> Knute Johnson wrote:
>> The volatiles exist because the methods that access them can be called
>> from
>> other threads. I could have synchronized the start() stop() methods
>> but not
>> easily the socket variable in the run() method. I thought it was
>> cleaner to
>> just use volatile.
>
> I see a problem right there.
>
>> public void disconnect() {
>> if (isConnected())
>> if (socket != null)
>> try {
>> socket.close();
>> } catch (IOException ioe) {
>> ioe.printStackTrace();
>> }
>> }
>
> Since these are controlled by separate synchronization (different
> 'volatile' variables) there's a race condition trying to work with both
> at once.

I don't think so.

> Also, 'socket' can become 'null' between the check for not 'null' and
> the 'close()' call.

Only if I set it to null and I don't.  That is there so that if the 
Socket doesn't make the first connection when disconnect() is called 
that I won't get a NPE.

> You need to synchronize with 'synchronized' or other strong mechanism.

I don't think so and here's why; isConnected only gets modified by one 
thread and read by another, socket is only modified by one thread and 
read by another.  In the disconnect() method, as soon as isConnected() 
is called, isConnected the volatile variable is read and that would make 
socket current even if it weren't volatile which it is but I didn't want 
to rely on side effects in case I changed code somewhere.

Anyway, disconnect() isn't getting called in this situation so it's not 
causing my problem.

I really appreciate everybody looking at this.  I've got a couple of 
ideas of where to code some traps and I'll have to put those in one 
night and see what happens.

-- 

Knute Johnson