[Navigation - nestednav] "ViewModelStore should be set before setGraph call" crash [177825470]

Fixed

Bug

Status Update

No update yet.

Description

ja...@google.com

created issue #1

Jan 18, 2021 02:10PM

Jetpack Compose release version: alpha10
Android Studio Build: AF4

This happens when using nested navigation

val navController = rememberNavController()
NavHost(navController = navController) {
navigation {
something(navController) {
}}}

and in NavGraphBuilder.something:

composable{
onCommit(onboardingComplete) {
if (!onboardingComplete) {
navController.navigate(MainDestinations.ONBOARDING_ROUTE)
}
}}

Commenting this line out yields no crashes and navController.navigate() is used in many other places.

Not sure if it has to do with nested nav, but it worked fine with two nested navhosts.

I've uploaded a fork of Owl where this happens:

https://github.com/android/compose-samples/compare/main...JoseAlcerreca:owl_nestednav_crash?expand=1

Crash in: Owl/app/src/main/java/com/example/owl/ui/courses/Courses.kt

Comments

il...@google.com <il...@google.com> Jan 19, 2021 06:07PM

Reassigned to jb...@google.com.

jb...@google.com <jb...@google.com> #2Jan 26, 2021 11:24PM

Reassigned to an...@google.com.

Project: platform/frameworks/support
Branch: androidx-master-dev

commit c60f33e229e31ab328ef6b59dab63b264954831c
Author: Alexandre Elias <aelias@google.com>
Date: Fri Jul 10 16:23:09 2020

Semantics no-op cleanups

Partly in response to lmr's broad code review, I did a pass of
superficial API/implementation cleanup. The main changes are:

- I changed each Boolean SemanticsProperty where false is equivalent to
not being present to take "Unit" instead. This is conceptually
clearer: it avoids questions like "can I cancel out a semantics from a
merged child by setting it to false?" Because "property = Unit" looks
weird, I also changed the style of these to "property()".

- I moved the Semantics id generator closer to where it's used, in
SemanticsModifierCore. I made it internal and an AtomicInt.
(Note that integer ids are heavily used in the Android
AccessibilityNodeInfo APIs so I can't simply remove them entirely.)

- I deleted dead code. Some examples include SemanticsHintOverrides,
a public API not connected to anything, and SemanticsPropertyKey
merge() open method which is never called. (In both cases I have
a different plan in mind for accessibility.)

Fixes: 145951226
Fixes: 145955412
Test: existing tests
Relnote: "Single-value semantics properties now use a calling style.
For example, 'semantics { hidden = true }' is now written as:
'semantics { hidden() }'."

Change-Id: Ic1afd12ea22c926babc9662f1804d80b33aa0cfc

M ui/integration-tests/benchmark/src/androidTest/java/androidx/ui/benchmark/test/LayoutNodeModifierBenchmark.kt
M ui/ui-core/api/0.1.0-dev15.txt
M ui/ui-core/api/current.txt
M ui/ui-core/api/public_plus_experimental_0.1.0-dev15.txt
M ui/ui-core/api/public_plus_experimental_current.txt
M ui/ui-core/api/restricted_0.1.0-dev15.txt
M ui/ui-core/api/restricted_current.txt
M ui/ui-core/src/androidAndroidTest/kotlin/androidx/ui/graphics/vector/VectorTest.kt
M ui/ui-core/src/androidAndroidTest/kotlin/androidx/ui/semantics/SemanticsTests.kt
M ui/ui-core/src/androidMain/kotlin/androidx/ui/core/AndroidActuals.kt
M ui/ui-core/src/androidMain/kotlin/androidx/ui/core/AndroidComposeView.kt
M ui/ui-core/src/androidMain/kotlin/androidx/ui/core/AndroidComposeViewAccessibilityDelegateCompat.kt
M ui/ui-core/src/androidMain/kotlin/androidx/ui/core/AndroidPopup.kt
M ui/ui-core/src/commonMain/kotlin/androidx/ui/core/Expect.kt
M ui/ui-core/src/commonMain/kotlin/androidx/ui/core/semantics/SemanticsConfiguration.kt
D ui/ui-core/src/commonMain/kotlin/androidx/ui/core/semantics/SemanticsHintOverrides.kt
M ui/ui-core/src/commonMain/kotlin/androidx/ui/core/semantics/SemanticsModifier.kt
M ui/ui-core/src/commonMain/kotlin/androidx/ui/core/semantics/SemanticsNode.kt
M ui/ui-core/src/commonMain/kotlin/androidx/ui/core/semantics/SemanticsOwner.kt
M ui/ui-core/src/commonMain/kotlin/androidx/ui/core/semantics/SemanticsWrapper.kt
M ui/ui-core/src/commonMain/kotlin/androidx/ui/semantics/SemanticsProperties.kt
M ui/ui-foundation/api/0.1.0-dev15.txt
M ui/ui-foundation/api/current.txt
M ui/ui-foundation/api/public_plus_experimental_0.1.0-dev15.txt
M ui/ui-foundation/api/public_plus_experimental_current.txt
M ui/ui-foundation/api/restricted_0.1.0-dev15.txt
M ui/ui-foundation/api/restricted_current.txt
M ui/ui-foundation/src/main/java/androidx/ui/foundation/Clickable.kt
M ui/ui-foundation/src/main/java/androidx/ui/foundation/Dialog.kt
M ui/ui-foundation/src/main/java/androidx/ui/foundation/Scroller.kt
M ui/ui-foundation/src/main/java/androidx/ui/foundation/selection/Selectable.kt
M ui/ui-foundation/src/main/java/androidx/ui/foundation/selection/Toggleable.kt
M ui/ui-foundation/src/main/java/androidx/ui/foundation/semantics/FoundationSemanticsProperties.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/ButtonTest.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/CardTest.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/CheckboxScreenshotTest.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/RadioButtonScreenshotTest.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/ScaffoldTest.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/SnackbarTest.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/SurfaceTest.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/ripple/RippleIndicationTest.kt
M ui/ui-material/src/androidTest/java/androidx/ui/material/textfield/TextFieldScreenshotTest.kt
M ui/ui-material/src/main/java/androidx/ui/material/AppBar.kt
M ui/ui-material/src/main/java/androidx/ui/material/TextFieldImpl.kt
M ui/ui-test/src/androidTest/java/androidx/ui/test/AssertsTest.kt
M ui/ui-test/src/androidTest/java/androidx/ui/test/CallSemanticsActionTest.kt
M ui/ui-test/src/androidTest/java/androidx/ui/test/ErrorMessagesTest.kt
M ui/ui-test/src/androidTest/java/androidx/ui/test/FindersTest.kt
M ui/ui-test/src/androidTest/java/androidx/ui/test/PrintToStringTest.kt
M ui/ui-test/src/androidTest/java/androidx/ui/test/ScrollToTest.kt
M ui/ui-test/src/androidTest/java/androidx/ui/test/TextActionsTest.kt
M ui/ui-test/src/main/java/androidx/ui/test/Actions.kt
M ui/ui-test/src/main/java/androidx/ui/test/Filters.kt
M ui/ui-text/api/0.1.0-dev15.txt
M ui/ui-text/api/current.txt
M ui/ui-text/api/public_plus_experimental_0.1.0-dev15.txt
M ui/ui-text/api/public_plus_experimental_current.txt
M ui/ui-text/api/restricted_0.1.0-dev15.txt
M ui/ui-text/api/restricted_current.txt
M ui/ui-text/src/commonMain/kotlin/androidx/ui/text/CoreTextField.kt
M ui/ui-text/src/commonMain/kotlin/androidx/ui/text/TextSemanticsProperties.kt

https://android-review.googlesource.com/1360099

an...@google.com <an...@google.com> #3Jan 27, 2021 10:51AM

Reassigned to jb...@google.com.

onCommit to DisposableEffect is mostly a rename, so I don't expect something to change in this regard

I also don't understand the relation between navController, lifecycleOwner, viewModelStore and onCommit - the param on onCommit is onboardingComplete, so this block should be restart when this changes

jb...@google.com <jb...@google.com> #4Jan 27, 2021 02:59PM

Reassigned to an...@google.com.

The DisposableEffect inside of NavHost is the one that is running a 2nd time with the same instances of navController, lifecycleOwner, viewModelStore. I thought if those instances are the same, that the block should be skipped.

an...@google.com <an...@google.com> #5Jan 27, 2021 03:08PM

Reassigned to jb...@google.com.

Are you sure they are the same objects? as far as I understand Jose is using nested nav controllers and calls navigate on the topmost one. this means we change the current navigation entry on the top and this changes the values for AmbientLifecycleOwner.current and AmbientViewModelStoreOwner.current.viewModelStore for the inner nav controller to the new entry object. Could you please verify this idea?

jb...@google.com <jb...@google.com> #6Jan 27, 2021 05:35PM

Reassigned to an...@google.com.

It looks that way. I attached screen shots from the first and second calls to the onCommit() scope and the objects are the same for both passes.

Owl Nested Nav onCommit 1.png

1.1 MB

View

Download

Owl Nested Nav onCommit 2.png

1.1 MB

View

Download

co...@gmail.com <co...@gmail.com> #7Jan 28, 2021 01:08AM

I've run into this issue and it happens when NavHost is recomposed but where you have the line calling rememberNavController() isn't recomposed.

I think the best solution is to move the navController set up code into rememberNavController() and make NavHost() just render the contents.

For me, this became an issue using Scaffold(). I'm trying to disable the drawer by setting drawerContent to null but this causes it to recompose and the NavHost in the body causes it to crash.

an...@google.com <an...@google.com> #8Jan 29, 2021 07:37PM

Reassigned to jb...@google.com.

Jeremy, in this case probably between two onCommit calls there is one onDispose call. so the composable is being disposed(the screen is not visible anymore?) and then attached again

ja...@google.com <ja...@google.com> #9Feb 22, 2021 03:19PM

Checked in beta snapshot and this is still an issue. Any updates or workarounds?

jb...@google.com <jb...@google.com> #10Feb 26, 2021 12:45AM

Currently working on a potential fix in aosp/1591051. Do you have an updated version of the app that uses the latest compose? If so I could check if that fix addresses your use case.

As far as work arounds, to avoid the crash you can clear the back stack when the NavHost is disposed by adding

onDispose {
  while (navController.popBackStack()) { }
}

ap...@google.com <ap...@google.com> #11Mar 1, 2021 09:22PM

Project: platform/frameworks/support
Branch: androidx-main

commit e2280407cf218dd031a8f74dba2b4d709fc3e826
Author: Jeremy Woods <jbwoods@google.com>
Date: Mon Mar 01 11:43:16 2021

No-op for same ViewModelStore and LifecycleOwner

When attempting to setViewModelStore or setLifecycleOwner if the
parameter passed in is the same as the one already set on the
NavController, the call should be a no-op.

RelNote: "Calling setViewModelStore or setLifecycleOwner with the same
object that has already been set will now be a no-op"
Test: NavControllerTest
Bug: 177825470

Change-Id: Idf491fa68c0771dd617832cefe4c03e8ea85dae7

M navigation/navigation-runtime/src/androidTest/java/androidx/navigation/NavControllerTest.kt
M navigation/navigation-runtime/src/main/java/androidx/navigation/NavController.java

https://android-review.googlesource.com/1611781

jb...@google.com <jb...@google.com> #12Mar 2, 2021 04:17PM

Marked as fixed.

This has been fixed internally and will be released as part of the Navigation 2.3.4 version.

You can verify this by following the snapshot instructions with build id 7177928 and switching your navigation dependency to 2.4.0-SNAPSHOT. Note that this will need to be an explicit dependency on navigation-runtime in addition to the dependency on navigation-compose.

ap...@google.com <ap...@google.com> #13Mar 2, 2021 06:07PM

Project: platform/frameworks/support
Branch: androidx-main

commit b76fc769718910bb2630c5e8c951d50df38e4d58
Author: Jeremy Woods <jbwoods@google.com>
Date: Tue Mar 02 09:22:58 2021

Bump nav-compose nav dependency to 2.3.4

Making nav-compose depend on navigation-runtime 2.3.4 for a fix to no-op
when setting the same ViewModelStore.

RelNote: "Navigation Compose now depends on Navigation version 2.3.4
which includes the fix for attempting to set the same ViewModelStore
object after the graph has been set."
Test: NavHostTest
Bug: 177825470

Change-Id: I65c24fcd5674c1ac72715c612854aac62c2eda88

M navigation/navigation-compose/build.gradle
M navigation/navigation-compose/src/androidTest/java/androidx/navigation/compose/NavHostTest.kt

https://android-review.googlesource.com/1612802

jb...@google.com <jb...@google.com> #14Mar 3, 2021 07:13PM

Now that Navigation compose depends on navigation-runtime 2.3.4, you can use the navigation-compose 1.0.0-SNAPSHOT with buildId 7179952.

Issue 177825470

Description

Issue summary

Comments

il...@google.com <il...@google.com> Jan 19, 2021 06:07PM

jb...@google.com <jb...@google.com> #2Jan 26, 2021 11:24PM

an...@google.com <an...@google.com> #3Jan 27, 2021 10:51AM

jb...@google.com <jb...@google.com> #4Jan 27, 2021 02:59PM

an...@google.com <an...@google.com> #5Jan 27, 2021 03:08PM

jb...@google.com <jb...@google.com> #6Jan 27, 2021 05:35PM

co...@gmail.com <co...@gmail.com> #7Jan 28, 2021 01:08AM

an...@google.com <an...@google.com> #8Jan 29, 2021 07:37PM

ja...@google.com <ja...@google.com> #9Feb 22, 2021 03:19PM

jb...@google.com <jb...@google.com> #10Feb 26, 2021 12:45AM

ap...@google.com <ap...@google.com> #11Mar 1, 2021 09:22PM

jb...@google.com <jb...@google.com> #12Mar 2, 2021 04:17PM

ap...@google.com <ap...@google.com> #13Mar 2, 2021 06:07PM

jb...@google.com <jb...@google.com> #14Mar 3, 2021 07:13PM

Add comment

Issue metadata