Sorry, I phrased that poorly.
Old scheduler with "old events" (i.e. from JSNI that correctly uses $entry, buggy JSNI forgetting $entry doesnt count) will work properly, and old/new events with new scheduler will work too. The problem is in old scheduler with "new" events - anything using plain jsinterop and avoiding $entry on its way into JS from Java. There is no fix we can do in GWT for it, old Scheduler assumes always that Entry owns all chances for JS to call Java, and so correctly wires up GWT's side of the event loop as well as uncaught exception handling.
Another option than triggering a fixed delay is to decide when the finally should be invoked, and call SchedulerImpl.INSTANCE.flushFinallyCommands(), that will trigger them all to take place right away. This is what $entry does, after your event is entirely handled in Java.
As a workaround, it may be possible to extend com.google.gwt.core.client.SchedulerImpl and splice in the Promise wiring found in gwt-core. Three bad effects off the top of my head:
* scheduleEntry won't work (deprecated in gwt-core, there is no way to do this without requiring $entry or the like)
* legacy dev mode won't work (it _might_ be possible to work around this, but for this last issue...)
* any browser without Promise support won't work, or will have a still-wrong hack in it (without promises, you need another microtask implementation, and of the browsers that don't have Promise, most don't have MutationObserver either which is the go-to hack I've found. Are there other hacks?) - you may end up with the "finally" happening after the next event or after some other timer instead of before it.
Filing it sounds good - we can offer the tradeoffs, and if someone has the need and the bandwidth to resolve this problem fully, we probably can get most of the way there, provided they can accept the caveats and test heavily.
--
Colin Alworth
colin@colinalworth.com
On Wed, Jun 24, 2020, at 5:09 PM, Gordan Krešić wrote:
> I'm willing to provide more info, but TBH I'm not sure what are you asking.
>
> You can see below code example that shows the problem. One "solution" that
> triggers scheduleFinally to run is to invoke another scheduleFixedDelay -
> that event is for sure not related to widgets, but I'm not sure if it is
> related to JsInterop.
>
> OTOH, I can confirm that handling mouse click on button widget DO triggers
> scheduleFinally, too.
>
> Like I've said, I've noticed new Sheduler (gwt-core) and all the differences
> that it brings when compared to "old" Scheduler (gwt-client), but I didn't
> test this new Scheduler.
>
> So, "old" Scheduler obviously has a bug or at least very unexpected
> behaviour. I agree that fixing this is not worth an effort if it's not a
> very simple fix (which I assume it isn't), but I offered to document this
> misbehaviour in form of bug report so that anyone else can at least find
> this workaround if needed (I didn't found any reference to it so far).
>
>
> On 24. 06. 2020. 19:16, Colin Alworth wrote:
> > Can you clarify the browser event that you are working with? If it is
> > something through JsInterop, then this is expected, since JSNI style calls
> > into Java from JS require $entry, but jsinterop provides no such mechanism.
> > If it is an event from a GWT Widget, then that would go through JSNI, and
> > this should not happen.
> >
> > As such, https://github.com/gwtproject/gwt-core has an updated Scheduler
> > which is compatible with JsInterop calls - scheduleFinally is now
> > implemented as a microtask, so that it will always precede any subsequent
> > event, no matter how the event originates. The gwt-core project also
> > contains an updated StyleInjector, and a corresponding gwt-resources module
> > (not yet moved to github.com/gwtproject) will provide an updated
> > ClientBundle implementation.
> >
> > Note that as we're committed to maintaining legacy dev mode in the GWT 2.x
> > branch, we cannot replace the com.google.gwt.core.client.Scheduler wiring
> > with jsinterop.
> >
> > On Sunday, June 21, 2020 at 6:13:54 AM UTC-5, Gordan Krešić wrote:
> >
> > I'll take the blame for code design, no questions about that :)
> >
> > Do you think it's worth filing a bug report or this is too much of an edge
> > case in a code that is going away soon anyway? Note that I didn't test
> > against org.gwtproject.core.client.Scheduler only
> > com.google.gwt.core.client.Scheduler (and they do differ, at least
> > regarding
> > scheduleFinally).
> >
> >
> > On 19. 06. 2020. 09:42, Thomas Broyer wrote:
> > > Yes, we can probably consider that a bug in GWT.
> > > I'd also call this pattern of doing real work in a static initializer
> > a code
> > > smell:
> > >
> > http://misko.hevery.com/code-reviewers-guide/flaw-constructor-does-real-work/
> > <http://misko.hevery.com/code-reviewers-guide/flaw-constructor-does-real-work/>
> >
> > > While still a flaw considering the above link, it's however a common
> > > practice to call `ensureInjected()` from the class constructor (ideally,
> > > you'd rather call it in a lifecycle method, such as Activity#start, or
> > > Widget#onAttach); and that would likely fix your issue here.
> > >
> > > On Thursday, June 18, 2020 at 8:33:06 PM UTC+2, Gordan Krešić wrote:
> > >
> > > On 18. 06. 2020. 20:25, Gordan Krešić wrote:
> > > > Probably unrelated with StyleInjector but with
> > Scheduler.scheduleFinally
> > >
> > > Ok, I've put a most basic repro case to prove that this is a
> > Scheduler
> > > issue:
> > >
> > > public class Foo {
> > >
> > > static {
> > > Scheduler.get().scheduleFinally(() ->
> > > GWT.log("Finally!"));
> > > }
> > >
> > > }
> > >
> > > Now initialite Foo on startup (in EntryPoint.onModuleLoad() for
> > example),
> > > but "Finally!" will be printed only *after* first event loop, like
> > > described:
> > >
> > > Scheduler.get().scheduleFixedDelay(() -> {
> > > return false;
> > > }, 0);
> > >
> > > -gkresic.
> > >
> > > --
> > > You received this message because you are subscribed to the Google
> > Groups
> > > "GWT Users" group.
> > > To unsubscribe from this group and stop receiving emails from it,
> > send an
> > > email to google-web-toolkit+unsubscribe@googlegroups.com
> > <mailto:google-web-toolkit%2Bunsubscribe@googlegroups.com>
> > > <mailto:google-web-toolkit+unsubscribe@googlegroups.com
> > <mailto:google-web-toolkit%2Bunsubscribe@googlegroups.com>>.
> > > To view this discussion on the web visit
> > >
> > https://groups.google.com/d/msgid/google-web-toolkit/1ecd7663-8a5a-449e-aef5-4008d3735433o%40googlegroups.com
> > <https://groups.google.com/d/msgid/google-web-toolkit/1ecd7663-8a5a-449e-aef5-4008d3735433o%40googlegroups.com>
> >
> > >
> > <https://groups.google.com/d/msgid/google-web-toolkit/1ecd7663-8a5a-449e-aef5-4008d3735433o%40googlegroups.com?utm_medium=email&utm_source=footer
> > <https://groups.google.com/d/msgid/google-web-toolkit/1ecd7663-8a5a-449e-aef5-4008d3735433o%40googlegroups.com?utm_medium=email&utm_source=footer>>.
> >
> >
> > --
> > You received this message because you are subscribed to the Google Groups
> > "GWT Users" group.
> > To unsubscribe from this group and stop receiving emails from it, send an
> > email to google-web-toolkit+unsubscribe@googlegroups.com
> > <mailto:google-web-toolkit+unsubscribe@googlegroups.com>.
> > To view this discussion on the web visit
> > https://groups.google.com/d/msgid/google-web-toolkit/0d761c71-bb99-44b2-9581-9aee45cff40bo%40googlegroups.com
> > <https://groups.google.com/d/msgid/google-web-toolkit/0d761c71-bb99-44b2-9581-9aee45cff40bo%40googlegroups.com?utm_medium=email&utm_source=footer>.
>
> --
> You received this message because you are subscribed to a topic in the
> Google Groups "GWT Users" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/google-web-toolkit/TyJAVyCkir4/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to
> google-web-toolkit+unsubscribe@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/google-web-toolkit/060994ea-9eef-9ecb-fa96-a54cceeac3d1%40steatoda.com.
>
--
You received this message because you are subscribed to the Google Groups "GWT Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to google-web-toolkit+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/google-web-toolkit/872580b3-db73-44f2-91d7-110a7e389d4a%40www.fastmail.com.
No comments:
Post a Comment