[rfc][icedtea-web] headless dialogues -tech preview
Andrew Azores
aazores at redhat.com
Wed Jun 17 14:39:22 UTC 2015
On 17/06/15 09:57 AM, Jiri Vanek wrote:
> On 06/17/2015 03:44 PM, Andrew Azores wrote:
>> On 17/06/15 08:47 AM, Jiri Vanek wrote:
>>> On 06/15/2015 06:01 PM, Andrew Azores wrote:
>>>> On 15/06/15 11:53 AM, Jiri Vanek wrote:
>>>>> On 06/15/2015 05:50 PM, Andrew Azores wrote:
>>>>>> On 15/06/15 11:48 AM, Jiri Vanek wrote:
>>>>>>> On 06/15/2015 05:36 PM, Andrew Azores wrote:
>>>>>>>> On 15/06/15 11:21 AM, Jiri Vanek wrote:
>>>>>>>>> On 06/15/2015 04:59 PM, Andrew Azores wrote:
>>>>>>>>>> Hi,
>>>>>>>>>>
>>>>>>>>>> On 14/06/15 02:04 PM, Jiri Vanek wrote:
>>>>>>>>>>> Hi!
>>>>>>> ...
>>>>>>>>> and - most important - headless ITW will not be run by average
>>>>>>>>> user :)
>>>>>>>>
>>>>>>>> I really think it's worthwhile to take the "better message"
>>>>>>>> approach rather than directly
>>>>>>>> presenting
>>>>>>>> the user with a stacktrace. If you wish then you can proceed
>>>>>>>> how you are now and I will work on
>>>>>>>> improving this later.
>>>>>>>
>>>>>>> I do not promise I will improve it in some close future.But I
>>>>>>> currently don't have simple
>>>>>>> solution
>>>>>>> for (definitely best) "help+suggested values, pass/better message"
>>>>>>>
>>>>>>> If you came up with anything, please do! If I wil comep with
>>>>>>> anything. I will definitively fix!
>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>> Type 'exit' to terminate ITW one f below values:
>>>>>>>> ...
>>>>>>>>>> blocker for this patch.
>>>>>>>>>
>>>>>>>>> This is not bug, its an feature!-)
>>>>>>>>>
>>>>>>>>> You can see how this text is composed? Also do you realize
>>>>>>>>> target audience of this patch? Also
>>>>>>>>> have you seen big comment " //this is default SecurityDialog
>>>>>>>>> "toString". All extending
>>>>>>>>> panels are
>>>>>>>>> recommanded to override " ?
>>>>>>>>>
>>>>>>>>> So this method is providing some *good* default text. (in v2
>>>>>>>>> the removal of html tags was much
>>>>>>>>> improved.
>>>>>>>>
>>>>>>>> Yes, the HTML tag removal does look much better now, although
>>>>>>>> I'd still personally like to
>>>>>>>> see it
>>>>>>>> cleaned up further. Still, this is better now. But I do think
>>>>>>>> that duplicating everything is not
>>>>>>>> very nice. Sure the target audience is not the average end user
>>>>>>>> but "YesYes\nNoNo" is going to
>>>>>>>> give
>>>>>>>> the more advanced user target audience the impression that this
>>>>>>>> feature wasn't really polished
>>>>>>>> before release, no?
>>>>>>>
>>>>>>> Does it need to be? The headless dialogues have to be making
>>>>>>> working again, or -headless have to
>>>>>>> be dropped. As I'm wring below. Feel free to override my stub
>>>>>>> method below with proper texts. I'm
>>>>>>> not going to do so. The changeset is well prepared for making it
>>>>>>> nicer.
>>>>>>>>
>>>>>>>>>
>>>>>>>>> The providing of texts for exactly those messages will be work
>>>>>>>>> for another changesets. MAybe
>>>>>>>>> for
>>>>>>>>> another person...
>>>>>>>>>
>>>>>>>>> Well - the headless dialogues are now completly broken, so
>>>>>>>>> *anything* is better then current
>>>>>>>>> state.
>>>>>>>>
>>>>>>>> True, but if something's worth doing then it's worth doing
>>>>>>>> correctly, right?
>>>>>>>
>>>>>>> Form point of this patch (which adds logic) it is really small
>>>>>>> nitpick to provide better texts.
>>>>>>> Feel free to add them on your own as another changeset...
>>>>>>>
>>>>>>> Or wait if somebody will help with it.
>>>>>>>
>>>>>>> Some of ITW's less used partsare in terrible shape. I need to
>>>>>>> make tehm work, so they canbe
>>>>>>> maintained without confusions. Once those are done. I will be
>>>>>>> finally free to quit.
>>>>>>>
>>>>>>> So I do not wont to provide cool headless dialogues. I wont to
>>>>>>> remove duplicate code, and make ti
>>>>>>> work.
>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Also, here's another bug for you:
>>>>>>>>>>
>>>>>>>>>>> Type 'exit' to terminate ITW one f below values:
>>>>>>>>>>> [YES, NO]
>>>>>>>>>>> netx: Initialization Error: Could not initialize applet.
>>>>>>>>>>> (Your Extended applets security
>>>>>>>>>>> is at
>>>>>>>>>
>>>>>>>>> yes I'm aware of it. I fixed it in v2. Bad typo on last second
>>>>>>>>> before posting :(
>>>>>>>>
>>>>>>>> "Type exit to terminate ITW one of below values" still doesn't
>>>>>>>> quite make sense.
>>>>>>>>
>>>>>>>> "Type exit to terminate ITW, or type one of the below values"
>>>>>>>> would be a fixed-up version of the
>>>>>>>> same message you have now.
>>>>>>>
>>>>>>>
>>>>>>> Type "exit" to terminate ITW, or type one of the below values
>>>>>>>
>>>>>>> Sure! Thanx!
>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>>>> 'high' and this application is missing the 'permissions'
>>>>>>>>>>> attribute in manifest. And you have
>>>>>>>>>>> refused to run it.)
>>>>>>>>>>> netx: Initialization Error: Could not initialize applet.
>>>>>>>>>>> (Your Extended applets security
>>>>>>>>>>> is at
>>>>>>>>>>> 'high' and this application is missing the 'permissions'
>>>>>>>>>>> attribute in manifest. And you have
>>>>>>>>>>> refused to run it.)
>>>>>>>>>>> net.sourceforge.jnlp.LaunchException: Fatal: Initialization
>>>>>>>>>>> Error: Could not initialize
>>>>>>>>>>> applet.
>>>>>>>>>>> For more information click "more information button".
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.Launcher.createApplet(Launcher.java:764)
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.Launcher.launchApplet(Launcher.java:653)
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.Launcher$TgThread.run(Launcher.java:939)
>>>>>>>>>>> Caused by: net.sourceforge.jnlp.LaunchException: Your
>>>>>>>>>>> Extended applets security is at 'high'
>>>>>>>>>>> and
>>>>>>>>>>> this application is missing the 'permissions' attribute in
>>>>>>>>>>> manifest. And you have refused to
>>>>>>>>>>> run
>>>>>>>>>>> it.
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.runtime.ManifestAttributesChecker.checkPermissionsAttribute(ManifestAttributesChecker.java:279)
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.runtime.ManifestAttributesChecker.checkAll(ManifestAttributesChecker.java:114)
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.initializeResources(JNLPClassLoader.java:777)
>>>>>>>>>>>
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.<init>(JNLPClassLoader.java:285)
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.createInstance(JNLPClassLoader.java:357)
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:429)
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.runtime.JNLPClassLoader.getInstance(JNLPClassLoader.java:403)
>>>>>>>>>>> at
>>>>>>>>>>> net.sourceforge.jnlp.Launcher.createApplet(Launcher.java:729)
>>>>>>>>>>> ... 2 more
>>>>>>>>>>
>>>>>>>>>> I typed "Control-D"/EOF character for this. At this point it
>>>>>>>>>> seems ITW is hanging.
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Atatched v2 again.
>>>>>>>>
>>>>>>>> EOF char for input still hangs it.
>>>>>>>
>>>>>>> When? What? how? How may I reproduce it?
>>>>>>
>>>>>> Wait for the "type one of the below values" prompt and then press
>>>>>> Control-D.
>>>>>
>>>>> Oh Yah!
>>>>>
>>>>> Nice catch:) The "if s == null" is missing.
>>>>>
>>>>> Do you agree on return defaultNEgativeAnswer and unlock?
>>>>
>>>> Sounds about right. To me, Ctrl-D in a "shell" sort of program is
>>>> synonymous with an exit command,
>>>> but using negative answer here sounds fine, too.
>>>>
>>>
>>> So there is version III. It still tech preview, because:
>>> - it contains trustAll/none hunkk from JNLPClaslaoder [1]
>>> and
>>> - contains fix to that ctrl+d you spotted above [0]
>>>
>>> But it should have fixed all the issues (even the "help+suggested
>>> values, pass/better message" and
>>> better default generated message!) you brought up.
>>> I'm really happy that the patch is so small only 200 lines. Rest are
>>> small refatroings.
>>> And in addition to v2 it have support for remebering. Both "jsut for
>>> record" and "pernament". And
>>> it supports headless keystore password too...
>>>
>>>
>>> [0] All "Launcher" related hunks
>>> It appeared, that the EFO was handeld properly since
>>> beginning,however, for some reason, JnlpBoot
>>> have been calling
>>> - Launcher launcher = new Launcher(false);
>>> instead of
>>> + Launcher launcher = new Launcher(true);
>>>
>>> According tojavadochere:
>>> * @param exitOnFailure Exit if there is an error (usually
>>> default, but false when being used
>>> from the plugin)
>>>
>>> After digging in code I have not found an reason why false should be
>>> passed to constructor here...
>>> I changed it, tested and it seems ok.
>>> Also, it also fixed issue, when typying NO was not terminating
>>> correctly, but it did not cleared
>>> out, why *clicking* on NO *did* terminatd correctly...
>>>
>>> Strange....
>>>
>>>
>>> [1]
>>> diff -r 8430c4b53e40
>>> netx/net/sourceforge/jnlp/runtime/JNLPClassLoader.java
>>> --- a/netx/net/sourceforge/jnlp/runtime/JNLPClassLoader.java Tue Jun
>>> 16 17:29:17 2015 +0200
>>> +++ b/netx/net/sourceforge/jnlp/runtime/JNLPClassLoader.java Wed Jun
>>> 17 14:27:31 2015 +0200
>>> @@ -1059,16 +1059,6 @@
>>> * @throws LaunchException if the user does not approve every
>>> dialog prompt.
>>> */
>>> private void checkTrustWithUser() throws LaunchException {
>>> - if (JNLPRuntime.isTrustNone()) {
>>> - if (!securityDelegate.getRunInSandbox()) {
>>> - setRunInSandbox();
>>> - }
>>> - return;
>>> - }
>>> - if (JNLPRuntime.isTrustAll() ||
>>> securityDelegate.getRunInSandbox()) {
>>> - return;
>>> - }
>>> -
>>> if (getSigningState() == SigningState.FULL &&
>>> jcv.isFullySigned() &&
>>> !jcv.getAlreadyTrustPublisher()) {
>>> jcv.checkTrustWithUser(securityDelegate, file);
>>> }
>>>
>>> Will go in as part of "[rfc][icedtea-web] mov repeated
>>> trusAll/none/shouldPromt/isHeadless to one
>>> palce"
>>
>> Was V3 supposed to be attached with this mail?
>>
> crap. yes. done
Well the patch certainly looks nice, but I can't get it to apply cleanly
to HEAD and so I haven't tested it. Mainly I just want to see what the
headless "dialog" prompt looks like and how interacting with it looks/feels.
--
Thanks,
Andrew Azores
More information about the distro-pkg-dev
mailing list