Adding an IRequestCycleListener that put a new value into the session at the end of each request DOES fix the problem. As you noted, it fixed my stripped-down example. I have tested all of the symptoms I have mentioned in previous related posts and they are ALL fixed :>
Overriding our session class's equals() method to return false did NOT solve the problem in my stripped-down example and I did not investigate that approach any further. Thanks so much, Martin, for your help on this!! I can't properly express how grateful I am. I don't know how long it would have taken for me to track that down without your help, but there would have been a lot of unhappiness along the way, I am sure :) I have a fix that I can proceed with, but I am puzzled why other Wicket/AppEngine users do not see this problem. It seems like nearly every Wicket app on GAE should be affected. Is there any more investigation you can recommend that might shed light on that? I'd like to either (1) figure out what we've done to aggravate this problem and fix it...or (2) get an appropriate fix into Wicket or GAE-Initializer, if that is the right thing to do. Chris On Fri, May 15, 2015 at 9:24 AM, Martin Grigorov <mgrigo...@apache.org> wrote: > On Fri, May 15, 2015 at 4:16 PM, Christopher Merrill < > ch...@webperformance.com> wrote: > > > Wow, that makes sense - and I _think_ it explains the symptoms I've seen > > with other problems as well. I'll have to re-investigate those. > > > > The issue you mentioned is marked as resolved in 6.12 and the related > > issue (WICKET-5473) was fixed in 6.14. We are running 6.19. So I must be > > reading those descriptions wrong...or this problem is somehow different. > > > > You read it correct. It should be fixed. > I don't know which version of Jetty AppEngine uses and what kind of checks > it does. > I just remembered about the symptoms in WICKET-5390 and thought that it > could be the same issue. > > > > > > Anyway, I'll test your suggestion (or perhaps override > > WebSession.equals()?). > > > > Test all of them. > > > > > > Chris > > > > On Fri, May 15, 2015 at 3:50 AM, Martin Grigorov <mgrigo...@apache.org> > > wrote: > > > > > Hi, > > > > > > AppEngine uses Jetty and I remembered about > > > https://issues.apache.org/jira/browse/WICKET-5390. > > > Here the reporter explains that Jetty won't update its SessionStore if > it > > > believes the http session didn't change. > > > Wicket stores both o.a.w.Session and all pages rendered in the last > > request > > > as attributes in the http session. > > > If Jetty doesn't detect the change(s) in the http session it will not > > store > > > it and later will serve a http session with obsolete data inside. > > > By doing "getSession().setAttribute("blah", somethingRandom)" we try to > > > help Jetty realize that there is a change and make sure it saves the > > latest > > > state. > > > You can move this helper code in IRequestCycleListener#onEndRequest() > so > > it > > > is executed for every request. > > > If I am correct then this should workaround the problem. > > > > > > Martin Grigorov > > > Wicket Training and Consulting > > > https://twitter.com/mtgrigorov > > > > > > On Fri, May 15, 2015 at 12:00 AM, Christopher Merrill < > > > ch...@webperformance.com> wrote: > > > > > > > Oh, and what you probably also wanted to know: When I press refresh > on > > > the > > > > page, the message from the constructor does not appear in the logs > > again. > > > > Also nothing after pressing the button (which does nothing). > > > > > > > > On Thu, May 14, 2015 at 4:50 PM, Christopher Merrill < > > > > ch...@webperformance.com> wrote: > > > > > > > > > I added the getSession().setAttribute() into the page constructor. > > > When I > > > > > hit the page, I see two lines in the AppEngine logs. The first is > for > > > the > > > > > URL I put in the browser (/portal/pages/Test), which indicates a > 302 > > > > > response to the browser and the log entry also contains the debug > > > message > > > > > that I included along with the setAttribute() line. > > > > > > > > > > The next line in the logs is for URL /portal/pages/Test?1 and is a > > 200 > > > > > response with the page content. > > > > > > > > > > Does that help? > > > > > > > > > > Chris > > > > > > > > > > On Thu, May 14, 2015 at 4:42 PM, Christopher Merrill < > > > > > ch...@webperformance.com> wrote: > > > > > > > > > >> I added that line, along with > > > > >> Debug.log.out("the test attribute is: " + > > > > >> getSession().getAttribute("test")); > > > > >> in the onSubmit() method of the AjaxButton. > > > > >> > > > > >> If I DO NOT refresh the page before pressing the button, then it > > > prints > > > > >> the message with the date in the logs. > > > > >> > > > > >> If I do refresh the page, nothing...which I take to mean that the > > code > > > > >> never runs. > > > > >> > > > > >> I'll try it in the constructor for the page. > > > > >> > > > > >> Chris > > > > >> > > > > >> > > > > >> On Thu, May 14, 2015 at 4:35 PM, Martin Grigorov < > > > mgrigo...@apache.org> > > > > >> wrote: > > > > >> > > > > >>> OK, then also put the same line in some constructor that is > > invoked. > > > > >>> > > > > >>> Martin Grigorov > > > > >>> Wicket Training and Consulting > > > > >>> https://twitter.com/mtgrigorov > > > > >>> > > > > >>> On Thu, May 14, 2015 at 11:32 PM, Christopher Merrill < > > > > >>> ch...@webperformance.com> wrote: > > > > >>> > > > > >>> > I'll do that. It'll take a few minutes to deploy that back up > to > > > > >>> AppEngine. > > > > >>> > > > > > >>> > But I don't think that code will be executed - I feel pretty > > > > confident > > > > >>> that > > > > >>> > the event listeners are not being invoked. At least, my debug > > > > >>> statements in > > > > >>> > the event listener do not get into the AppEngine logs. > > > > >>> > > > > > >>> > Chris > > > > >>> > > > > > >>> > > > > > >>> > On Thu, May 14, 2015 at 4:23 PM, Martin Grigorov < > > > > mgrigo...@apache.org > > > > >>> > > > > > >>> > wrote: > > > > >>> > > > > > >>> > > Can you try something else: > > > > >>> > > in onClick() add code like: getSession.setAttribute("test", > new > > > > >>> Date()) > > > > >>> > > > > > > >>> > > Martin Grigorov > > > > >>> > > Wicket Training and Consulting > > > > >>> > > https://twitter.com/mtgrigorov > > > > >>> > > > > > > >>> > > On Thu, May 14, 2015 at 11:18 PM, Christopher Merrill < > > > > >>> > > ch...@webperformance.com> wrote: > > > > >>> > > > > > > >>> > > > Thanks for your help, Martin! > > > > >>> > > > > > > > >>> > > > I turned off page recreation by adding: > > > > >>> > > > > > > > >>> > > > > getPageSettings().setRecreateMountedPagesAfterExpiry(false); > > > > >>> > > > > > > > >>> > > > to my Application.init() method. > > > > >>> > > > > > > > >>> > > > There is no change in the behavior, either locally or in > > > > AppEngine > > > > >>> -- > > > > >>> > > i.e. > > > > >>> > > > no page expired errors. > > > > >>> > > > > > > > >>> > > > You can observe the behavior here: > > > > >>> > > > http://1.wp-portal-staging.appspot.com/portal/pages/Test > > > > >>> > > > > > > > >>> > > > by visiting the page, refreshing it and then pressing a > > button > > > > >>> (which > > > > >>> > > will > > > > >>> > > > do nothing if you refreshed). I posted the code for that > page > > > in > > > > >>> > another > > > > >>> > > > thread ("Model value change lost between > > AjaxButton.onSubmit() > > > > and > > > > >>> > > > form.onSubmit()"), in case that helps. > > > > >>> > > > > > > > >>> > > > TIA! > > > > >>> > > > Chris > > > > >>> > > > > > > > >>> > > > > > > > >>> > > > > > > > >>> > > > On Thu, May 14, 2015 at 3:43 PM, Martin Grigorov < > > > > >>> mgrigo...@apache.org > > > > >>> > > > > > > >>> > > > wrote: > > > > >>> > > > > > > > >>> > > > > Hi, > > > > >>> > > > > > > > > >>> > > > > My guess is that there is some problem with the saving of > > the > > > > >>> page in > > > > >>> > > the > > > > >>> > > > > backing stores. > > > > >>> > > > > Try by disabling page recreation for expired pages, see > > > > >>> > IPageSettings, > > > > >>> > > > > If I'm right then you should start seeing PageExpiredPage > > > after > > > > >>> > > clicking > > > > >>> > > > > the link. > > > > >>> > > > > The next step is will be to find why the pages could not > be > > > > >>> stored. > > > > >>> > > > > > > > > >>> > > > > Martin Grigorov > > > > >>> > > > > Wicket Training and Consulting > > > > >>> > > > > https://twitter.com/mtgrigorov > > > > >>> > > > > > > > > >>> > > > > On Thu, May 14, 2015 at 9:40 PM, Christopher Merrill < > > > > >>> > > > > ch...@webperformance.com> wrote: > > > > >>> > > > > > > > > >>> > > > > > I'm seeing problems throughout our app where refreshing > > the > > > > >>> page > > > > >>> > > causes > > > > >>> > > > > > other event listeners to then malfunction. These are > > > > trivially > > > > >>> > simple > > > > >>> > > > > > listeners, like: > > > > >>> > > > > > > > > > >>> > > > > > Link customer_link = new Link("org_link") > > > > >>> > > > > > { > > > > >>> > > > > > @Override > > > > >>> > > > > > public void onClick() > > > > >>> > > > > > { > > > > >>> > > > > > setResponsePage(new > > > > >>> > > > > > OrganizationAdministrationPage(getPageReference(), > > > > >>> > > > > > _organization_key)); > > > > >>> > > > > > } > > > > >>> > > > > > > > > > >>> > > > > > private static final long serialVersionUID = > > > > >>> > > -6396556709778260098L; > > > > >>> > > > > > }; > > > > >>> > > > > > > > > > >>> > > > > > > > > > >>> > > > > > Instead of returning the response page, the browser > > > receives > > > > a > > > > >>> > > forward > > > > >>> > > > > > (302) back to the same page instance...which of course > > > means > > > > >>> that > > > > >>> > the > > > > >>> > > > > link > > > > >>> > > > > > appears to do nothing. > > > > >>> > > > > > > > > > >>> > > > > > I can only reproduce this behavior when the app is > > deployed > > > > to > > > > >>> > > > AppEngine > > > > >>> > > > > -- > > > > >>> > > > > > it works fine running on my desktop. The problems > started > > > > when > > > > >>> we > > > > >>> > > > > upgraded > > > > >>> > > > > > the application to Wicket 6. I'm not implying this is a > > > > Wicket > > > > >>> 6 > > > > >>> > > > > > problem...in fact I assume we have done something that > is > > > > >>> causing > > > > >>> > > this > > > > >>> > > > > > malfunction. But since it worked under Wicket 5, I'm > > hoping > > > > >>> someone > > > > >>> > > > will > > > > >>> > > > > > see a relationship between the symptoms and a change > that > > > we > > > > >>> need > > > > >>> > to > > > > >>> > > > make > > > > >>> > > > > > for compatibility with Wicket 6. > > > > >>> > > > > > > > > > >>> > > > > > Any hints or debugging ideas (since I can't step > through > > > the > > > > >>> code > > > > >>> > > > running > > > > >>> > > > > > in AppEngine) would be greatly appreciated! > > > > >>> > > > > > > > > > >>> > > > > > Chris > > > > >>> > > > > > > > > > >>> > > > > > > > > >>> > > > > > > > >>> > > > > > > >>> > > > > > >>> > > > > >> > > > > >> > > > > > > > > > > > > > > >
