[rfc][icedtea-web] offline support

[Jiri Vanek]

On 08/05/2014 05:04 PM, Andrew Azores wrote:
> On 08/05/2014 11:01 AM, Jiri Vanek wrote:
>> On 08/05/2014 04:50 PM, Andrew Azores wrote:
>>> On 08/04/2014 11:34 AM, Jiri Vanek wrote:
>>>> +    public static boolean isOnline() {
>>>> +        if (isOfflineForced()) {
>>>> +            return !isOfflineForced();
>>>> +        }
>>>> +        return onlineDetected;
>>>> +    }
>>>
>>> This hunk is a little bit silly IMO :) that first return can just be "return false".
>>
>> will be fixed:
>>
>>  public static boolean isOnline() {
>>         if (isOfflineForced()) {
>>             return false;
>>         }
>>         return isOnlineDetected();
>>     }
>>
>> (there was hidden NPE)
>>
>>>
>>> I'm also not a fan (and I'm sure you're not surprised here) of onlineDetected being a Boolean, but
>>> since the null value is only ever used in internal state and all the outward facing accessors use
>>> plain boolean, I guess it's okay.
>>
>> I thought you will be complaining and I had this internall usage as ace in sleeve.
>>>
>>> Otherwise... I can't find much else to complain about. It seems to work fine and I generally like
>>> the patch. There is one thing I'm wondering about though. Using this patch with a large application
>>> (in this case Elluminate), the application still takes quite some time to start up. Is this simply
>>> because of how many local JARs there are to verify? Or maybe ResourceTracker is still sending HEAD
>>> requests to try to check if the JARs are up to date... ?
>>
>> Yes. If you enable debug/show console. You wills see that *all* this time is consumed by
>> cerverifier.  Once offline is detected, no attempts  to  verify  cahce are done. It should not
>> touch network at all. However, there are few url.openConnections in ITW about which usages I'm not
>> sure what to do. I will investigate them later and try to replace them (during testing, I have not
>> seen them used when implementing this patch)
>>
>> I have run Elluminate during testing:)
>>
>> I have jsut found this method:
>>
>> /**
>>      * Return true if the Environment is Offline
>>      */
>>     @Override
>>     public boolean isOffline() {
>>
>>         URL url = findFirstURLFromJNLPFile();
>>         URLConnection conn = null;
>>         try {
>>             conn = url.openConnection();
>>             conn.getInputStream().close();
>>             return false;
>>         } catch (IOException exception) {
>>             return true;
>>         } finally {
>>             if (conn != null && conn instanceof HttpURLConnection) {
>>                 ((HttpURLConnection) conn).disconnect();
>>             }
>>         }
>>     }
>>
>>
>> I will shorten it via new is offline detection:
>>
>>
>>     @Override
>>     public boolean isOffline() {
>>
>>         URL url = findFirstURLFromJNLPFile();
>>         JNLPRuntime.detectOnline(url);
>>         return !JNLPRuntime.isOnline();
>>     }
>>
>>
>> Ok for head? I will prepare 1.5 version (much shorter) later today.
>>
>>
>> J.
>
> All looks/sounds good to me, ok for HEAD.
>
> Thanks,
>

Here is 1.5 version. Head's one did not applied at all so I wrote it again. I hope I did not messed it.
In this version the Xservice is untouched, Xoflfine is missing and so allowOflfine can not be avoided.
The forceOffline was kept for debuging purposes and is missing  the setter

Otherwise it *should* be same.

J.

Once this is (no rush!) approved, I would like to freeze branch and sent tarball to you and Lukas 
and Jie for testing.  Andrew, although you are officially leaving 8th, will you be able to lead them 
? I'm afraid the burden of 1.5 testing is completely in Toronto this time.

JJ.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: offlineSupport_15_3.patch
Type: text/x-patch
Size: 13531 bytes
Desc: not available
URL: <http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20140806/f40dec24/offlineSupport_15_3.patch>