Clojurians Log v2

Gonna fire this up now....

Confirmed the issue. Oh, btw, it should not matter (it's a NOP), but we only need to wrap state change in with-cc in watches, when a state change is already underway. So the gesture-detector :onTap is fine. But I have those commented out in the elevated button onPresses and the problem is unchanged. I will leave them commented out while I explore, just to KISS. I suspect I may be missing a setState. Deep-diving...

Confirm it works without the gesture-detector -- good clue! Just realizing how strange is the behavior -- the text hi/howdy proves we are changing columns! Btw, eliminating the "but does it?" button altogether works in the sense that "you'd think" disappears. Good one! Wondering if we are looking at two issues... 🤔

OK, the second column runs as expected, and as we can see from the text being pulled in. I have to crash, but methinks Flutter thinks the two GDs (gesture detectors) are the same, for some value of "same". More after a quick crash.

Sounds likely! Some bits in this thread are perhaps relevant: https://clojurians.slack.com/archives/C03A6GE8D32/p1675818632237609

Hmm, maybe it has something to do with GuestureDetector being a StatelessWidget ?

https://clojurians.slack.com/archives/C03A6GE8D32/p1675887308660429?thread_ts=1675818632.237609&cid=C03A6GE8D32

Since we're not dealing with a stateful widget, maybe we need to call markNeedsBuild directly?

Thx, I have been getting ready to explore which widgets were stateless and which not. I think I have experimented enough with widget juggling, I'll see where I can work markNeedsBuild into things.

Ooh, wait, this might be a very simple fix, this deftag looks suspicious...

(deftag tiltontec.flutter-mx.factory/k1-child-stateful gesture-detector m/GestureDetector)

@hiskennyness Aha, that's the culprit! swapped with k1-child-stateless and now it works like a charm :)

Great catch, and the Flutter doc confirms I got that wrong. Now I have to cheak Ink, because a center/ink/text is also broken, no GD at all.

Not sure about Ink itself, but InkWell is stateless, so probably same for Ink? I wish flutter gave us a simpler way to find out :P

Oh, yeah, InkWell is indeed stateful, (in accordence with the deftag), so something else must be at play.

Ink seems to be either. No inheritance (I scroll waaayyyyyyyy down to an inheritance list) and see neither Stateful nor Stateless.

Oh you're right, I got my pages mixed up for a sec.

ISTR Text is the same.

Hmm, (deftag tiltontec.flutter-mx.factory/k1-child-stateless ink m/Ink) work?

With text I covered all the bases:

(deftagleaf tiltontec.flutter-mx.factory/konly-param1-stateless text m/Text)
(deftagleaf tiltontec.flutter-mx.factory/konly-param1-stateful text! m/Text)

And ISTR that was necessitated by sth not updating.

The Joy of Clojure Flutter :P

Maybe taking a more direct approach of markNeedsBuild instead of setState might work around the flutter dance?

Would need some extra guards in the mx internals, I'd guess, but might be a universal solution.

(I really have little context to work with here, this is just me stabbing in the dark)

So far I have found Flutter and Dart to be a PITA to get to work, but eventually sort it out in a way that makes sense. Then I get to look at mobile, desktop, and Web. 🙂 And f/mx can hide the nuisance. With other tools, the pain never ends, we just soldier on grumbling. Are you able to toggle back and forth OK. What if you hit the first button twice, then hit the second? I am seeing:

The following assertion was thrown while handling a gesture:
md-awaken> incoming md-state not= :nascent, it is:md-quiesced

> Are you able to toggle back and forth OK. What if you hit the first button twice, then hit the second? Hmm, yes this works without incident on my end. Granted I've patched state-set awhile ago because my bluetooth callbacks were trying to set state while things were still building. Not sure if that why or not, but just in case it's relevant. :

(defn state-set
  ([^State state]
   (state-set state nil))
  ([^State state me]
   (if (.-mounted state)
     (do (dpx :setting-state state (minfo me))
         (try (.setState state (fn [] (do)))
              (catch Exception e
                (Future.delayed Duration.zero
                                (fn [] (if (.-mounted state)) (.setState state (fn [] (do))))))
              #_ (finally (Future.delayed Duration.zero
                                (fn [] (if (.-mounted state)) (.setState state (fn [] (do))))))))
     #_(dp :NOT_SETTING_STATE_NOT_MOUNTED state (cty/minfo me)))))

Ah, I had jazzed things up a little:

(button (str "but does it?" (* 10 (mget (fasc :scaffo) :step)))
  (fn [] (mx/dp :button-2)))

Not able to recreate. Meanwhile, I broke the x03-physics-sim example which expects the GD to have state. Hopefully just need to work in that "nearest state" utility:

(cF (when-let [st (fx/my-state)]
      (m/AnimationController .vsync st)))

So I changed the physics sim to search up to the nearest state, after making GD stateless, and got an error saying it was not a TickerProvider. WTF? A code search and I discover that, in desperation a year ago, I had splatted this line on exactly one factory variant, k1-child-stateful:

^:mixin m/SingleTickerProviderStateMixin ;; todo make optional

Hmm, works fine without the fx/guesture-detector .