[rfc] [icedtea-web] better error reporting for LaunchException in plugin
Adam Domurad
adomurad at redhat.com
Wed Jan 30 06:58:01 PST 2013
On 01/30/2013 07:42 AM, Jiri Vanek wrote:
> On 01/29/2013 05:01 PM, Adam Domurad wrote:
>> On 01/21/2013 06:35 AM, Jiri Vanek wrote:
>>> Launch errors are thrown from netx, and printed out .. somewhere. At
>>> the end plugin (correctly)
>>> fails with NPE.
>>> It Is not nice. Eg in SplashError dialogue they can not be shown.
>>>
>>> I have added static accumulator for LaunchErrors (== all errors
>>> preceding npe in plugin) and then
>>> I'm adding this chain of exception to Splash error report.
>>>
>>> It is quite useful - see example - but there is one pitfall. As
>>> accumulator being static, it is
>>> cumulating errors from all running applets :(
>>> However I have not found a way how to distinguish between individual
>>> applets :( (any idea welcomed!)
>>
>> Hmm. Well each applet has an associated thread-group -- you could use
>> that to determine the current
>> applet.
>
> Tried.. Do not work :((
>
> The best will be to know jnlpFile(or pluginbridge) and then have map
> jnlpFile,LaunchError - however, see below.
>>
>> As well, I do not like adding a static accumulator to each
>> constructor -- I would prefer a special
>
> it is not so nasty :)
>> launch error utility class that is always used, that takes a JNLPFile
>> (in cases where its available)
>> -or- tries to determine one by the current ThreadGroup. This will
>> unfortunately need some additional
>> 'infrastructure'.
>
> Yap, then there is question if such a infrastructure and following
> refactoring is worthy - IMHO is not:(
>
> So just mior fixed patch attached.
>>
>>> I have added small warkaround with timestamps and localised
>>> explanation.. and.. it is still
>>> better then nothing :)
>>>
>>>
>>> J.
>>
>> I know you want this in as 'better than nothing' but it does seem too
>> hackish to me :-/ IMO
>> constructors should not have such side-effects.
>
> It is not side effect. This is quite common.
> I agree it is hackich. But not so bad as you consider it for :))
Agree to disagree :-)
>
>> However I am all for LaunchException's capturing a time-stamp in
>> their constructor.
>>
>>>
>>>
>>> **example**
>>>
>>> So now instead of plain, nothing saying
>>> "IcedTea-Web Plugin version: 1.4pre+rb646c8b9c2e2+
>>> Mon Jan 21 12:29:17 CET 2013
>>> java.lang.NullPointerException
>>> at net.sourceforge.jnlp.NetxPanel.runLoader(NetxPanel.java:130)
>>> at sun.applet.AppletPanel.run(AppletPanel.java:380)
>>> at java.lang.Thread.run(Thread.java:679)"
>>>
>>> Is there:
>>>
>>> IcedTea-Web Plugin version: 1.4pre+rb646c8b9c2e2+
>>> Mon Jan 21 12:29:17 CET 2013
>>> java.lang.NullPointerException
>>> at net.sourceforge.jnlp.NetxPanel.runLoader(NetxPanel.java:130)
>>> at sun.applet.AppletPanel.run(AppletPanel.java:380)
>>> at java.lang.Thread.run(Thread.java:679)
>>>
>>> Chain:
>>> 1) at Mon Jan 21 12:27:27 CET 2013
>>> net.sourceforge.jnlp.LaunchException: Fatal: Application Error: The
>>> applet was unsigned. The
>>> applet was unsigned, and the security policy prevented it from running.
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.initializeResources(JNLPClassLoader.java:680)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.<init>(JNLPClassLoader.java:240)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:374)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:347)
>>> at net.sourceforge.jnlp.Launcher.createApplet(Launcher.java:751)
>>> at net.sourceforge.jnlp.Launcher.getApplet(Launcher.java:713)
>>> at net.sourceforge.jnlp.Launcher$TgThread.run(Launcher.java:942)
>>> 2) at Mon Jan 21 12:27:27 CET 2013
>>> net.sourceforge.jnlp.LaunchException: Fatal: Initialization Error:
>>> Could not initialize applet.
>>> at net.sourceforge.jnlp.Launcher.createApplet(Launcher.java:784)
>>> at net.sourceforge.jnlp.Launcher.getApplet(Launcher.java:713)
>>> at net.sourceforge.jnlp.Launcher$TgThread.run(Launcher.java:942)
>>> Caused by: net.sourceforge.jnlp.LaunchException: Fatal: Application
>>> Error: The applet was
>>> unsigned. The applet was unsigned, and the security policy prevented
>>> it from running.
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.initializeResources(JNLPClassLoader.java:680)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.<init>(JNLPClassLoader.java:240)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:374)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:347)
>>> at net.sourceforge.jnlp.Launcher.createApplet(Launcher.java:751)
>>> ... 2 more
>>> 3) at Mon Jan 21 12:27:29 CET 2013
>>> net.sourceforge.jnlp.LaunchException: Fatal: Application Error: The
>>> applet was unsigned. The
>>> applet was unsigned, and the security policy prevented it from running.
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.initializeResources(JNLPClassLoader.java:680)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.<init>(JNLPClassLoader.java:240)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:374)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:347)
>>> at net.sourceforge.jnlp.Launcher.createApplet(Launcher.java:751)
>>> at net.sourceforge.jnlp.Launcher.getApplet(Launcher.java:713)
>>> at net.sourceforge.jnlp.Launcher$TgThread.run(Launcher.java:942)
>>> 4) at Mon Jan 21 12:27:29 CET 2013
>>> net.sourceforge.jnlp.LaunchException: Fatal: Initialization Error:
>>> Could not initialize applet.
>>> at net.sourceforge.jnlp.Launcher.createApplet(Launcher.java:784)
>>> at net.sourceforge.jnlp.Launcher.getApplet(Launcher.java:713)
>>> at net.sourceforge.jnlp.Launcher$TgThread.run(Launcher.java:942)
>>> Caused by: net.sourceforge.jnlp.LaunchException: Fatal: Application
>>> Error: The applet was
>>> unsigned. The applet was unsigned, and the security policy prevented
>>> it from running.
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.initializeResources(JNLPClassLoader.java:680)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.<init>(JNLPClassLoader.java:240)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:374)
>>> at
>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:347)
>>> at net.sourceforge.jnlp.Launcher.createApplet(Launcher.java:751)
>>> ... 2 more
>>>
>>
>> Comments inline.
>>
>>
>>> diff -r b646c8b9c2e2 netx/net/sourceforge/jnlp/LaunchException.java
>>> --- a/netx/net/sourceforge/jnlp/LaunchException.java Fri Jan 18
>>> 12:49:37 2013 +0100
>>> +++ b/netx/net/sourceforge/jnlp/LaunchException.java Mon Jan 21
>>> 12:20:20 2013 +0100
>>> @@ -16,6 +16,10 @@
>>>
>>> package net.sourceforge.jnlp;
>>>
>>> +import java.util.Date;
>>> +import java.util.LinkedList;
>>> +import java.util.List;
>>> +
>>> /**
>>> * Thrown when a JNLP application, applet, or installer could not
>>> * be created.
>>> @@ -25,6 +29,29 @@
>>> */
>>> public class LaunchException extends Exception {
>>>
>>> +
>>> + public static class LaunchExceptionWithStamp{
>>> + private final LaunchException ex;
>>> + private final Date stamp;
>>> +
>>> + private LaunchExceptionWithStamp(LaunchException ex) {
>>> + this.ex=ex;
>>> + this.stamp=new Date();
>>> + }
>>> +
>>> + public LaunchException getEx() {
>>> + return ex;
>>> + }
>>> +
>>> + public Date getStamp() {
>>> + return stamp;
>>> + }
>>> +
>>> +
>>> +
>>> + }
>>> + public static final List<LaunchExceptionWithStamp>
>>> launchExceptionChain = new
>>> LinkedList<LaunchExceptionWithStamp>();
>>> +
>>> private static final long serialVersionUID = 7283827853612357423L;
>>>
>>> /** the file being launched */
>>> @@ -54,6 +81,7 @@
>>> this.summary = summary;
>>> this.description = description;
>>> this.severity = severity;
>>> + saveLaunchException(this);
>>
>> Bit too much of a hack, for me.\
> it is really not so evil:)
>>
>>> }
>>>
>>> /**
>>> @@ -61,6 +89,7 @@
>>> */
>>> public LaunchException(Throwable cause) {
>>> super(cause);
>>> + saveLaunchException(this);
>>> }
>>>
>>> /**
>>> @@ -68,6 +97,7 @@
>>> */
>>> public LaunchException(String message, Throwable cause) {
>>> super(message, cause);
>>> + saveLaunchException(this);
>>> }
>>>
>>> /**
>>> @@ -78,6 +108,7 @@
>>> */
>>> public LaunchException(String message) {
>>> super(message);
>>> + saveLaunchException(this);
>>> }
>>>
>>> /**
>>> @@ -117,4 +148,9 @@
>>> return severity;
>>> }
>>>
>>> + private void saveLaunchException(LaunchException ex) {
>>
>> Should be synchronized, surely ?
> agree
>
>>
>>> + launchExceptionChain.add(new LaunchExceptionWithStamp(ex));
>>> +
>>> + }
>>> +
>>> }
>>> diff -r b646c8b9c2e2
>>> netx/net/sourceforge/jnlp/resources/Messages.properties
>>> --- a/netx/net/sourceforge/jnlp/resources/Messages.properties Fri
>>> Jan 18 12:49:37 2013 +0100
>>> +++ b/netx/net/sourceforge/jnlp/resources/Messages.properties Mon
>>> Jan 21 12:20:20 2013 +0100
>>> @@ -478,3 +478,4 @@
>>> SPLASHdefaultHomepage = Unspecified homepage, verify source rather
>>> SPLASHerrorInInformation = Error during loading of information
>>> element, verify source rather
>>> SPLASHmissingInformation = Information element is missing, verify
>>> source rather
>>> +SPLASHchainWas = There is list of exceptions which provided start
>>> of you applet. Please note,
>>> those exceptions can be from multiple applets. For good bug report,
>>> it is wothy to run your applet
>>> isolated.
>>
>> Hm, s/wothy/worthy/, but probably better is:
>> 'This is the list of exceptions that occurred launching your applet'
>> ... 'For a good bug report, be
>> sure to run only one applet.'
>
> As always, As you wish.
>>
>>> diff -r b646c8b9c2e2
>>> netx/net/sourceforge/jnlp/resources/Messages_cs_CZ.properties
>>> --- a/netx/net/sourceforge/jnlp/resources/Messages_cs_CZ.properties
>>> Fri Jan 18 12:49:37 2013 +0100
>>> +++ b/netx/net/sourceforge/jnlp/resources/Messages_cs_CZ.properties
>>> Mon Jan 21 12:20:20 2013 +0100
>>> @@ -468,4 +468,5 @@
>>> SPLASHanotherInfo= Dal\u0161\u00ed dostupn\u00e9 informace
> snip
>
> J.
>
OK. Improving it significantly did seem difficult.
Just a thought though -- I think the exception launch lists shouldn't be
public, I'm thinking access to it needs to be synchronized. Should
probably synchronize + return copy.
Once that is fixed, OK for HEAD.
-Adam
More information about the distro-pkg-dev
mailing list