Arrow keys scroll the page with SDL2

[iboB]

As specified in the docs the canvas is supposed to take the focus (unless I specify SDL_EMSCRIPTEN_KEYBOARD_ELEMENT). While it does correctly capture backspace for example, the arrow keys are still sent to the browser, which causes scrolling of the page when the intention is to send commands to the js application.

Is there a way to stop sending the arrow key events to the browser?

[kripken]

Do you have a testcase showing the issue? It works when I try e.g. tests/sdl2_text.c, the page doesn't scroll with the arrow keys.

[iboB]

Here is a small repro program: https://github.com/iboB/ibob.github.com/blob/master/tests/emscripten/sdl.cpp

You can see it here running too: ibob.github.io/tests/emscripten/sdl.html
This is compiled in Windows with the line: > emcc sdl.cpp -o sdl.html -O2 -std=c++11 -s USE_SDL=2

When I open the html in Google Chrome or MS IE 11 in a small-enough window, pressing the arrow keys will also scroll the page

[iboB]

This cannot be reproduced in Firefox

[kripken]

I can confirm, I see scrolling in chrome but not in firefox.

No idea which browser is right here. We should probably make a tiny (non-compiled) testcase and file issues on the browser vendors.

[kripken]

I filed emscripten-ports/SDL2#21 , let's continue the discussion there as this is an SDL2 issue.

[sbc100]

Reopening since https://github.com/emscripten-ports/SDL2/ was archived. Unless we want to open an upstream SDL issue on the main SDL repo this seems like the best place to continue discussion.

[rtrussell]

Unless we want to open an upstream SDL issue

I can't see anything obviously wrong with the code in SDL_emscriptenevents.c, which still seems to check the navigation keys:

        SDL_bool prevent_default = SDL_TRUE;
        ....
        if ((scancode == SDL_SCANCODE_BACKSPACE) ||
        (scancode == SDL_SCANCODE_TAB) ||
        (scancode == SDL_SCANCODE_LEFT) ||
        (scancode == SDL_SCANCODE_UP) ||
        (scancode == SDL_SCANCODE_RIGHT) ||
        (scancode == SDL_SCANCODE_DOWN) ||
        ((scancode >= SDL_SCANCODE_F1) && (scancode <= SDL_SCANCODE_F15)) ||
        keyEvent->ctrlKey) {
        is_nav_key = SDL_TRUE;
    }

    if ((eventType == EMSCRIPTEN_EVENT_KEYDOWN) && SDL_EventEnabled(SDL_EVENT_TEXT_INPUT) && !is_nav_key) {
        prevent_default = SDL_FALSE;
    }

    return prevent_default;

What is seemingly supposed to happen is that if the callback function passed to emscripten_set_keydown_callback() returns true the native (canvas) action of the key is suppressed. It appears that the above code does return true for the navigation keys.

[sbc100]

Does the suggested fix looks right to you: emscripten-ports/SDL2#21 (comment)

We have a feature where you can try EMCC_LOCAL_PORTS="sdl2=/home/alon/Dev/ports/SDL2" to try out your own fork of SDL2 if you want to try it out. Otherwise mabye @Daft-Freak can pick it up again?

[rtrussell]

Does the suggested fix looks right to you: emscripten-ports/SDL2#21 (comment)

That is the code already in SDL2 (listed above), isn't it? That was my point, superficially SDL looks to be correct, so is the problem that Emscripten is not acting properly on the return value?

[sbc100]

I don't know I was just going by the last comment left by @Daft-Freak: "Looks like I never actually applied my fix for arrow keys. I'll try to look at this again soon." Perhaps it landed but never got tested. I let them respond.