Releasing Alt sometimes inserts characters

[dmchurch]

Issue Type: Bug

I don't have a reliable STR, unfortunately, but I've been having an intermittent issue for quite a while which manifests as follows:

1. Hold the Alt key. (Left Alt, if you're curious.)
2. Press Up/Down to move the current line/selection up or down
3. Release the Alt key.

The editor then acts as if I'd typed a character; sometimes R, or X, or sometimes an undisplayable Unicode character, among others. This means it gets inserted at the cursor point, or if I had a selection, it replaces the selection.

Of course, now that I'm trying to reproduce it, it doesn't happen, so I don't expect anyone else can track it down for me; for all I know it may be some weird interaction with OS-level software. What I want to know is, does VSC have some sort of internal logging where I can see what actions it has performed recently, and why? If I could narrow down what it thought it was doing, I could maybe figure out why it was doing it.

VS Code version: Code 1.63.2 (899d46d, 2021-12-15T09:40:02.816Z)
OS version: Windows_NT x64 10.0.22000
Restricted Mode: No

System Info

Item	Value
CPUs	Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz (12 x 2208)
GPU Status	2d_canvas: enabled gpu_compositing: enabled multiple_raster_threads: enabled_on oop_rasterization: enabled opengl: enabled_on rasterization: enabled skia_renderer: enabled_on video_decode: enabled vulkan: disabled_off webgl: enabled webgl2: enabled
Load (avg)	undefined
Memory (System)	31.92GB (18.28GB free)
Process Argv	--enable-proposed-api redhat.java --crash-reporter-id ac355797-2567-4ffb-9b4c-ea2f04db70d0
Screen Reader	no
VM	0%

Extensions (23)

Extension	Author (truncated)	Version
java-decompiler	Ben	0.0.3
save-editors-layout	ctf	0.2.2
remove-multi-cursors	dmc	0.0.1
gitlens	eam	11.7.0
remotehub	Git	0.22.0
latex-workshop	Jam	8.23.0
local-code	loc	0.0.4
remote-containers	ms-	0.209.6
remote-ssh	ms-	0.70.0
remote-ssh-edit	ms-	0.70.0
remote-wsl	ms-	0.63.13
vscode-remote-extensionpack	ms-	0.21.0
hexeditor	ms-	1.9.3
gradle-language	nac	0.2.3
java	red	1.2.0
functional-cursors	spo	0.0.8
vscodeintellicode	Vis	1.2.15
vscode-gradle	vsc	3.9.0
vscode-java-debug	vsc	0.37.0
vscode-java-dependency	vsc	0.18.9
vscode-java-pack	vsc	0.20.0
vscode-java-test	vsc	0.33.1
vscode-maven	vsc	0.34.2

A/B Experiments

vsliv368:30146709
vsreu685:30147344
python383:30185418
vspor879:30202332
vspor708:30202333
vspor363:30204092
pythonvspyl392cf:30425750
pythontb:30283811
pythonptprofiler:30281270
vshan820:30294714
vstes263cf:30335440
pythondataviewer:30285071
vscod805:30301674
pythonvspyt200:30340761
binariesv615:30325510
bridge0708:30335490
bridge0723:30353136
vsaa593cf:30376535
vsc1dsc:30424895
pythonvs932:30410667
vscop804:30404766
vscop341:30404997
vsrem710cf:30416617
vsbas813:30426126
vsic-wlc-tsjs:30426436

[alexdima]

@dmchurch What kind of keyboard are you on? Are perhaps the arrow keys sometimes acting as a numpad? The only special kind of input on Windows that I'm aware of that sounds similar would be the Alt+Numpad Keys special character insertion. Here's a site that explains this.

ALT+082 would create an R (maybe up arrow (8), down arrow (2) )
Also ALT+088 would create an X (maybe up arrow (8), up arrow (8) )

[dmchurch]

@alexdima !!!!!!!! omg you're absolutely right, that's totally what it is. Wow, I haven't thought about Alt keycodes in like two decades! And I absolutely do use the numpad on my laptop keyboard for navigation, since it doesn't have independent home/end/etc keys. (I end up switching numlock a lot when I'm on this laptop.)

Y'know, I'd thought that Windows only did Alt+numpad insertions when numlock was on, but clearly that's not the case. At least I have a workaround, though - using the dedicated arrow keys for alt+up/down!

It does bring up a question, though: is there any way to tell Windows not to process that key combination as part of an alt code if VSCode is handling it as an event?

[alexdima]

I'm glad that was the root cause. We can continue tracking in #22364 , but it doesn't seem likely that things will improve, I remember reporting this upstream, but nothing happening.