United States International Keyboard layout not working properly

[timea-techgirl]

I use US International Keyboard layout and for example, typing, "I'm" doesn't work correctly.

[rom1v]

"I'm" doesn't work correctly.

What's the result?

For me, it's work correctly.

[timea-techgirl]

"I'm" doesn't work correctly.

What's the result?

For me, it's work correctly.

This is the result. Im' or I 'm
It works correctly on Vysor.
Using windows 10. English - US International keyboard.

I press the letter I and single inverted comma and m = Im'
OR press the letter I and single inverted comma and space and m = I 'm

Both work perfectly in international keyboard. I have never had this issue in any app.

[rom1v]

Just to find the cause, do you still have the issue if you type very slowly? I, wait 2 seconds, ', wait 2 seconds, m?

[timea-techgirl]

Yes. I tried waiting more than 2 seconds. Still the same problem. My computer is very fast.

[rom1v]

What if you copy-paste I'm from the computer to scrcpy (instead of typing it)?

[timea-techgirl]

It copies correctly.

[rom1v]

English - US International keyboard.

Is this your layout on Windows or Android?

I added some logs:

diff --git a/server/src/main/java/com/genymobile/scrcpy/EventController.java b/server/src/main/java/com/genymobile/scrcpy/EventController.java
index 341869f..7ce96c1 100644
--- a/server/src/main/java/com/genymobile/scrcpy/EventController.java
+++ b/server/src/main/java/com/genymobile/scrcpy/EventController.java
@@ -97,6 +97,7 @@ public class EventController {
             return false;
         }
         for (KeyEvent event : events) {
+            Ln.d(" --> keyCode=" + event.getKeyCode() + ", action=" + event.getAction() + ", metaState=0x" + Integer.toHexString(event.getMetaState()));
             if (!injectEvent(event)) {
                 return false;
             }
@@ -105,6 +106,7 @@ public class EventController {
     }
 
     private boolean injectText(String text) {
+        Ln.d("injectText(\"" + text.replaceAll("\"", "\\\"") + "\")");
         for (char c : text.toCharArray()) {
             if (!injectChar(c)) {
                 return false;
@@ -144,6 +146,7 @@ public class EventController {
     }
 
     private boolean injectKeyEvent(int action, int keyCode, int repeat, int metaState) {
+        Ln.d("injectKeyEvent(" + action + ", " + keyCode + ", " + repeat + ", 0x" + Integer.toHexString(metaState) + ")");
         long now = SystemClock.uptimeMillis();
         KeyEvent event = new KeyEvent(now, now, action, keyCode, repeat, metaState, KeyCharacterMap.VIRTUAL_KEYBOARD, 0, 0,
                 InputDevice.SOURCE_KEYBOARD);
@@ -151,6 +154,7 @@ public class EventController {
     }
 
     private boolean injectKeycode(int keyCode) {
+        Ln.d("injectKeycode(" + keyCode + ")");
         return injectKeyEvent(KeyEvent.ACTION_DOWN, keyCode, 0, 0)
                 && injectKeyEvent(KeyEvent.ACTION_UP, keyCode, 0, 0);
     }

Here is a server built with these changes: scrcpy-server.jar
SHA-256: 29562c0558cac71675568c9cb6feb908d7ca377a95555b71d5d3c94a0b7afb90

Replace yours by this one. Then, in scrcpy, just press I, ' and m. It will print debug logs in your console. For me, it prints:

DEBUG: injectKeyEvent(0, 37, 0, 0x81)
DEBUG: injectKeyEvent(1, 37, 0, 0x0)
DEBUG: injectText("'")
DEBUG:  --> keyCode=75, action=0, metaState=0x0
DEBUG:  --> keyCode=75, action=1, metaState=0x0
DEBUG: injectKeyEvent(0, 41, 0, 0x0)
DEBUG: injectKeyEvent(1, 41, 0, 0x0)

What does it print for you?

[timea-techgirl]

I'm using US international in windows. I've been using it since I started using my computer.

DEBUG: injectKeyEvent(0, 37, 0, 0x41)
DEBUG: injectKeyEvent(1, 37, 0, 0x41)
DEBUG: injectKeyEvent(0, 41, 0, 0x0)
DEBUG: injectText("'")
DEBUG:  --> keyCode=75, action=0, metaState=0x0
DEBUG:  --> keyCode=75, action=1, metaState=0x0
DEBUG: injectKeyEvent(1, 41, 0, 0x0)

[rom1v]

Did you type slowly?

I will analyze this when I have some time. Thank you.

[timea-techgirl]

Yes, I did. I tried multiple times. And got the same output.

[rom1v]

Sorry, could I ask you to test with this additional patch, to get the logs in the order we receive it on the client side, please?

diff --git a/app/src/scrcpy.c b/app/src/scrcpy.c
index b777b77..d843810 100644
--- a/app/src/scrcpy.c
+++ b/app/src/scrcpy.c
@@ -147,10 +147,14 @@ handle_event(SDL_Event *event, bool control) {
             if (!control) {
                 break;
             }
+            LOGD("TEXTINPUT: [%s]\n", event->text.text);
             input_manager_process_text_input(&input_manager, &event->text);
             break;
         case SDL_KEYDOWN:
         case SDL_KEYUP:
+            LOGD("%s %x (%c)\n",
+                 event->type == SDL_KEYDOWN ? "KEYDOWN" : "KEYUP",
+                 (int) event->key.keysym.sym, (char) event->key.keysym.sym);
             // some key events do not interact with the device, so process the
             // event even if control is disabled
             input_manager_process_key(&input_manager, &event->key, control);

[timea-techgirl]

I don't understand what I need to do with this code. Did you miss uploading "scrcpy-server.jar" with this patch?

Please feel free to ask me anything any number of times. :)

[rom1v]

Here is a compiled version: scrcpy.exe.
SHA-256: 6f87290178123ab0d3abbf5a62269507a7ab876eb5a6b781707cf553c5bacdd5

[timea-techgirl]

scrcpy-server.jar: 1 file pushed. 1.7 MB/s (50495 bytes in 0.028s)
DEBUG: Starting stream thread
DEBUG: Starting controller thread
INFO: Created renderer: direct3d
INFO: Initial texture: 1080x1920
DEBUG: KEYDOWN 400000e1 (�)
DEBUG: KEYDOWN 69 (i)
DEBUG: TEXTINPUT: [I]
DEBUG: injectKeyEvent(0, 37, 0, 0x41)
DEBUG: KEYUP 69 (i)
DEBUG: injectKeyEvent(1, 37, 0, 0x41)
DEBUG: KEYUP 400000e1 (�)
DEBUG: KEYDOWN 27 (')
DEBUG: KEYUP 27 (')
DEBUG: KEYDOWN 6d (m)
DEBUG: TEXTINPUT: [']
DEBUG: TEXTINPUT: [m]
DEBUG: injectKeyEvent(0, 41, 0, 0x0)
DEBUG: injectText("'")
DEBUG:  --> keyCode=75, action=0, metaState=0x0
DEBUG:  --> keyCode=75, action=1, metaState=0x0
DEBUG: KEYUP 6d (m)
DEBUG: injectKeyEvent(1, 41, 0, 0x0)

scrcpy-server.jar: 1 file pushed. 1.7 MB/s (19734 bytes in 0.011s)
DEBUG: Starting stream thread
DEBUG: Starting controller thread
INFO: Created renderer: direct3d
INFO: Initial texture: 1080x1920
DEBUG: KEYDOWN 400000e1 (�)
DEBUG: KEYDOWN 400000e1 (�)
DEBUG: KEYDOWN 400000e1 (�)
DEBUG: KEYDOWN 400000e1 (�)
DEBUG: KEYDOWN 69 (i)
DEBUG: TEXTINPUT: [I]
DEBUG: KEYUP 69 (i)
DEBUG: KEYUP 400000e1 (�)
DEBUG: KEYDOWN 27 (')
DEBUG: KEYUP 27 (')
DEBUG: KEYDOWN 6d (m)
DEBUG: TEXTINPUT: [']
DEBUG: TEXTINPUT: [m]
DEBUG: KEYUP 6d (m)

[rom1v]

OK, thank you.

So when you press ', it waits for the next letter before generating the text input (which is weird, it means we cannot send ' (if we rely on text input) before you type m).

Maybe it triggers a SDL_TEXTEDITING event (https://wiki.libsdl.org/Tutorials/TextInput), but anyway, that will not solve the problem.

The result is messed up due to the workaround to fix bug #87: it sends letters as keys but other characters as text.

I wanted a "single mode" so that the user don't have to configure, but in the end there are several side-effects. I will probably add an option to select between "raw keys events", "text" (the behavior before v1.4) and "mixed" (the current behavior).

[timea-techgirl]

Oh I see. Thanks for the explanation. :)

[rom1v]

Does it still happen?

What if you pass --prefer-text?

[timea-techgirl]

It still happens.

It's fixed with this --prefer-text command.

[rom1v]

OK, thank you. 👍

This is "fixed" then (this is why --prefer-text had been implemented c916af0, because a single reasonable behavior could not work for all situations).