Godot: Android: LineEdit does not respond correctly to ENTER pressed on virtual keyboard

[mudbile]

Godot version:
Godot 3.2

OS/device including version:
Tested on Android 6.0

Issue description:
Pressing the ENTER button on the virtual keyboard while focused on a LineEdit node does not fire the "text_entered" signal. Instead, the first time you press ENTER it might change to a recommendation, and then actual newlines are appended (I think), resulting in abnormal graphical effects (I think either the node moves up or the view moves down). When you close the virtual keyboard the LineEdit reappears where it's supposed to be.

Steps to reproduce:

1. Create a project with a LineEdit and a Label
2. Connect the LineEdit's "text_entered" to a function that updates the text on the Label
3. Run on Android
4. Type something into the LineEdit and press the ENTER key
5. Observe that the Label does not update
6. Keep pressing ENTER
7. Observe weird graphical effects

Minimal reproduction project:
LineEditBug.zip

[capnm]

At least for us this is a big regression in the controls LineEdit or TextEdit.
We don't have such problems when we revert our projects to v3.2.1.stable. The ENTER key on the Android keyboard broken behavior makes a text input and editing under Android in v3.2.2.stable unusable :-(

[capnm]

Similar effects as described in #37190 also occur, or the TextEdit simply disappears after you press the Android ENTER key. And the virtual keyboard with the LineEdit now remains open :-(

[capnm]

Tested the official 3.2.3.rc2 and it is broken for reasons unknown to me:
the enter/return-key doesn't cause a new-line in the TextEdit and doesn't close the virtual Android keyboard in LineEdit.

Also AFAIK this:

• woks in 3.2.1
• is broken in 3.2.2
• works in 3.2.2 + https://github.com/capnm/godot/commit/13456724f56cf5dad2ddd852ae2fcdee7574e74c from #40469
• is broken in 3.2.3.rc2

[pouleyKetchoupp]

I can confirm #40469 & #40672 both seem to be missing in the official 3.2.3.rc2 release.

@akien-mga Is there an easy way to check that java changes have been included into the 3.2.3.rc2 android template?

[akien-mga]

It's built from ac2e7d87d1c398db9c796afba6973a2f170ddfa2, so the changes should be included.

I'll check logs to see if anything weird happened.

[akien-mga]

Welp no, it's built from basically the same state as RC 1, with some additional weirdness... :upside_down_face:

Damn :D

That's what happens when you do git pull origin/3.2 instead of git pull origin 3.2 :sob:

[akien-mga]

It's still uploading and the binaries will be fully online within ~30 min, but here's 3.2.3 RC 3: https://downloads.tuxfamily.org/godotengine/3.2.3/rc3/

This time properly built from 78a6a578843928017c1fe40f23a31a620e76d2ef.

@capnm @pouleyKetchoupp Could you test it as soon as possible and make sure that it works as intended? If it works, I'd likely publish 3.2.3-stable tomorrow or on Sunday (otherwise it will have to wait after my holidays so more mid-August :)).

[capnm]

The Godot executable thinks it's rc2, you probably didn't run the script mentioned in the README ...

[pouleyKetchoupp]

@akien-mga Apart from the editor still set as rc2, I can confirm the Android fixes work properly if I set the rc3 templates manually during export.

[capnm]

the Android fixes work properly if I set the rc3 templates manually during export.

Yeah, after editing the installed rc3 templates to rc2, it now works :+1:

[akien-mga]

Oh well... I need to improve that part of the build workflow to prevent these issues from happening...

I'll rebuild 3.2.3 RC 3 properly then, including a couple more fixes.

Thanks for confirming that the Android fixes work properly though, we should be good to go with 3.2.3-stable on that side.

[capnm]

@pouleyKetchoupp Unfortunately I found a new issue in the RC3 with your other PR that makes the adjustment fixes. The virtual touch-pad from Android now is/remains open if the virtual KB is closed and cuts off part of the Godot viewport window:

[pouleyKetchoupp]

@capnm I can reproduce it too. I'm checking if there's an easy fix for this bug.

[pouleyKetchoupp]

@akien-mga @capnm I've just opened #40944 to address the regression on 3.2.