False positive for MissingConstraints lint check with constraintlayout's Flow [181056543]

Verified

Bug

adexe s nau

Status Update

No update yet.

Description

tn...@google.com

created issue #1

Feb 23, 2021 10:20PM

With 4.2, given this layout (just replace activity_main in a default Empty project):

<androidx.constraintlayout.widget.ConstraintLayout xmlns:android="http://schemas.android.com/apk/res/android"
    xmlns:app="http://schemas.android.com/apk/res-auto"
    xmlns:tools="http://schemas.android.com/tools"
    android:layout_width="match_parent"
    android:layout_height="match_parent"
    tools:context=".MainActivity">

    <androidx.constraintlayout.helper.widget.Flow
        android:layout_width="wrap_content"
        android:layout_height="wrap_content"
        app:constraint_referenced_ids="someTextView"
        app:layout_constraintStart_toStartOf="parent"
        app:layout_constraintTop_toTopOf="parent" />

    <TextView
        android:id="@+id/someTextView"
        android:layout_width="wrap_content"
        android:layout_height="wrap_content"
        android:text="Hello World!" />

</androidx.constraintlayout.widget.ConstraintLayout>

and you run gradle :app:lintDebug on this layout, you'll get this incorrect error:

app/src/main/res/layout/activity_main.xml:16: Error: This view is not constrained. It only has designtime positions, so it will jump to (0,0) at runtime unless you add the constraints [MissingConstraints]
      <TextView
       ~~~~~~~~

Comments

tn...@google.com <tn...@google.com> #2Feb 23, 2021 10:25PM

The issue seems to be some discrepancy in what the IDE thinks the package name should be for the Flow, and what the compiled code says (see screenshot). Interestingly I only see this in 4.2, not 7.0, so perhaps these constants were updated recently.

Screen Shot 2021-02-23 at 2.13.29 PM.png

232 KB

View

Download

tn...@google.com <tn...@google.com> #3Feb 23, 2021 10:32PM

Diego, can you take a look at this?

I'm seeing something weird; when I create a lint unit test, the SdkConstants.CLASS_CONSTRAINT_LAYOUT_FLOW has the correct value for newName, so the test passes, but when I attach with a debugger to AGP 4.2, I get what you see in the screenshot -- the wrong value. Do you have any ideas what could cause the AndroidX name lookup to produce different values in these scenarios??

tn...@google.com <tn...@google.com> #4Feb 25, 2021 08:26PM

Reassigned to di...@google.com.

Haha, I meant to assign to Diego but forgot to actually do that.

In the meantime, for 4.2, I think I'll just check for both paths (pre AndroidX and post AndroidX).

di...@google.com <di...@google.com> #5Feb 25, 2021 08:40PM

Ack. I'll have a look.

tn...@google.com <tn...@google.com> #6Feb 25, 2021 09:02PM

(I have a 4.2-safe workaround in ag/13699079 if the root cause here is trickier to fix (especially given that this isn't happening in A/7.0, though it's possible that the same cause is there but hidden by something else.)

di...@google.com <di...@google.com> #7Feb 26, 2021 12:34PM

I did a bit of debugging and I think I know the problem.

com.android.tools.build.gradle:4.2.0-beta05 depends on com.android.tools.build.gradle:bundletool:1.1.0 and the later is shipping a very outdated version of migration.xml.

di...@google.com <di...@google.com> #8Feb 26, 2021 12:45PM

7.0.0-alpha04 has that dependency upgraded to bundletool:1.4.0 which would explain why it works there.

di...@google.com <di...@google.com> #9Feb 26, 2021 12:46PM

Reassigned to cm...@google.com.

Re-assigning to Chris just to double check that everything makes sense and the fact that we were are shipping 4.2.0 with quite an old version of bundletool.

cm...@google.com <cm...@google.com> #10Feb 26, 2021 02:31PM

Any duplicated classes should be renamed in bundle-tool, so there's something strange going on here

cm...@google.com <cm...@google.com> #11Feb 26, 2021 02:34PM

Ah I see that the migrateToAndroidx java resource isn't renamed...

cm...@google.com <cm...@google.com> #12Feb 26, 2021 02:38PM

Accepted by cm...@google.com.

Diego, can we (as a workaround for now) rename that java resource to avoid the conflict, and then I'll reach out to the bundle-tool team to rename that java resource

cm...@google.com <cm...@google.com> #13Feb 26, 2021 02:54PM

Assigned to di...@google.com.

Proposed fix is I7eacac3a70e30f6cf530fe68e6616dfebbb535e3

di...@google.com <di...@google.com> #14Feb 26, 2021 03:39PM

LGTM. Thanks Chris!

tn...@google.com <tn...@google.com> #15May 3, 2021 11:25PM

This is fixed right? (Chris' fix went into one of the last 4.2 betas.)

di...@google.com <di...@google.com> #16May 4, 2021 12:39PM

Verified by di...@google.com.

Verified in 4.2 RC1