Improve Emscripten support #857

cactorium · 2017-02-12T21:20:57Z

Mostly backporting winit functions into emscripten's platform module. It compiles and works against a modified version of badboy's glium example: https://github.com/cactorium/spring-crabs/tree/emscripten

tomaka · 2017-02-13T06:56:41Z

Err.. I don't understand what you're trying to achieve, but don't duplicate winit. Just continue importing it even on emscripten, but just don't call any of its functions.

cactorium · 2017-02-13T07:02:14Z

That breaks because winit doesn't support emscripten:

   Compiling winit v0.5.10
error[E0432]: unresolved import `self::platform::*`
 --> /home/kelvin/.cargo/registry/src/gh.neting.cc-1ecc6299db9ec823/winit-0.5.10/src/platform/mod.rs:1:9
  |
1 | pub use self::platform::*;
  |         ^^^^^^^^^^^^^^^^^^ Could not find `platform` in `platform`

error[E0432]: unresolved import `this_platform_is_not_supported`
  --> /home/kelvin/.cargo/registry/src/gh.neting.cc-1ecc6299db9ec823/winit-0.5.10/src/platform/mod.rs:22:5
   |
22 | use this_platform_is_not_supported;
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ no `this_platform_is_not_supported` in the root

# etc...

Like it's impossible to get it to compile for emscripten, unless there's some flag or trick that I don't know about?

tomaka · 2017-02-13T07:29:18Z

Winit should get a "null" backend which always returns an error when you try to use it.

cactorium · 2017-02-13T13:17:29Z

winit's got this thing in src/platform/mod.rs that's causing the error whenever you're targetting asmjs-unknown-emscripten:

#[cfg(all(not(target_os = "ios"), not(target_os = "windows"), not(target_os = "linux"),
  not(target_os = "macos"), not(target_os = "android"), not(target_os = "dragonfly"),
  not(target_os = "freebsd"), not(target_os = "openbsd")))]
use this_platform_is_not_supported;

A fresh pull of the master branch of winit fails to compile with target asmjs-unknown-emscripten:

[kelvin@batmanbatman repos]$ cd winit/
[kelvin@batmanbatman winit]$ git pull
Already up-to-date.
[kelvin@batmanbatman winit]$ cargo build --target asmjs-unknown-emscripten
warning: unused manifest key: package.categories
    Updating registry `https://github.com/rust-lang/crates.io-index`
   Compiling libc v0.2.20
   Compiling lazy_static v0.2.2
   Compiling shared_library v0.1.5
   Compiling winit v0.5.10 (file:///home/kelvin/repos/winit)
error[E0432]: unresolved import `self::platform::*`
 --> src/platform/mod.rs:1:9
  |
1 | pub use self::platform::*;
  |         ^^^^^^^^^^^^^^^^^^ Could not find `platform` in `platform`

error[E0432]: unresolved import `this_platform_is_not_supported`
  --> src/platform/mod.rs:22:5
   |
22 | use this_platform_is_not_supported;
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ no `this_platform_is_not_supported` in the root

error[E0433]: failed to resolve. Could not find `Window2` in `platform`
   --> src/window.rs:114:22
    |
114 |         let w = try!(platform::Window2::new(events_loop.events_loop.clone(), &self.window, &self.platform_specific));
    |                      ^^^^^^^^^^^^^^^^^^^^^^ Could not find `Window2` in `platform`

error[E0412]: type name `platform::MonitorId` is undefined or not in scope
   --> src/window.rs:306:24
    |
306 |     data: VecDequeIter<platform::MonitorId>,
    |                        ^^^^^^^^^^^^^^^^^^^ undefined or not in scope
    |
...

tomaka · 2017-02-13T13:37:43Z

Yes, what I mean is that winit should be modified to add a "null" implementation that is used instead of this "this_platform_is_not_supported" thing.

cactorium · 2017-02-13T13:47:34Z

Ohhh okay, sorry!

cactorium · 2017-02-13T14:48:31Z

So I think I got the default platform thing working, but I can't test because of the winit API transition thing; glutin doesn't compile with either emscripten or linux against the tip of winit, with or without my changes... Any recommendations on what to do next?

tomaka · 2017-02-13T15:10:49Z

Oh, I forgot about that.
You could do your changes against the dev-0.5 branch of winit.
This branch still uses the old API.

tomaka · 2017-02-13T16:05:13Z

src/api/emscripten/mod.rs

+               winit_builder: winit::WindowBuilder)
+                -> Result<Window, CreationError> {
+
+        let winit_window = winit_builder.build().unwrap();


The idea is that you wouldn't use winit at all. Just don't call it, and remove the winit_window field from the struct.

tomaka · 2017-02-13T16:05:38Z

src/api/emscripten/mod.rs

@@ -190,7 +161,7 @@ impl Window {

    #[inline]
    pub fn create_window_proxy(&self) -> WindowProxy {
-        WindowProxy
+        self.winit_window.create_window_proxy()


It's fine to panic! here. Nobody uses the window proxy, and it's going to be removed soon.

cactorium · 2017-02-13T16:39:47Z

Fixed both I think

cactorium · 2017-02-14T19:10:30Z

So I've got some work done on adding in the hooks for mouse/keyboard events as well, would those be good to add to this PR, or would opening a new PR after this one's merged be better?

tomaka · 2017-02-14T19:23:46Z

@cactorium As you wish.
Basically I wanted to publish winit 0.5.11 (which I forgot to do yesterday), then ask you to tweak the Cargo.toml of glutin in this PR to point to winit 0.5.11, then merge.

cactorium · 2017-02-15T05:50:17Z

So I guess I'll see how far I get before winit 0.5.11 gets published and submit whatever work I get done by then? Most of the event stuff should be working pretty solidly right now; the biggest hole left is converting to keys to VirtualKeyCodes

cactorium · 2017-02-15T14:38:37Z

All's ready except for the version change

gui1117 · 2017-02-15T15:12:55Z

I just want to mention that I started implementing emscripten plateform in winit there

I included keyboard but mouse because there was an issue with mousemove when the cursor is grabbed

tomaka · 2017-02-15T16:10:46Z

I'm hesitating between putting the emscripten stuff in glutin or in winit.

The initial reason why I wanted to stick to glutin is that you can only use OpenGL with emscripten anyway, and that the concept of a "window" is a bit blurry when it comes to web pages.
That being said Android has the same problem where you're manipulating a window but the whole screen.

But that's a pretty weak argument, so I'm not sure.

cactorium · 2017-02-16T15:40:01Z

I kind of agree with the idea of having it in glutin because of the fact that you don't really control a window. On the other hand, I think emscripten supports using SDL instead of OpenGL, so maybe it'd make sense to put event processing stuff in winit to share it across the two? On Wed, Feb 15, 2017, 11:10 AM tomaka <notifications@github.com> wrote: I'm hesitating between putting the emscripten stuff in glutin or in winit. The initial reason why I wanted to stick to glutin is that you can only use OpenGL with emscripten anyway, and that the concept of a "window" is a bit blurry when it comes to web pages. That being said Android has the same problem where you're manipulating a window but the whole screen. But that's a pretty weak argument, so I'm not sure. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#857 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABcW37Qyj51Xj16U9yroFC1NCnxxbjNgks5rcyOGgaJpZM4L-nvg> .

cactorium · 2017-02-17T10:26:38Z

My bad, I missed the version bump on winit. Cargo.toml's updated

grab mode use pointerlock https://www.w3.org/TR/pointerlock/ this behavior is not the same as other backend. I wanted this because as set_cursor_position is not available on emscripten backend. the only way to have a FPS camera is to use pointerlock. So you can use grab to have FPS camera on emscripten. also show_mouse is implement in javascript as in a pull request that hasn't been merged emscripten-core/emscripten#4616

more implementations

two fixes: * when mouse is grabbed send relative mousemove event * when mouse grabbed is refused, ask again.

Implement touch + fix mouse grab + catch more errors

gui1117 · 2017-05-03T10:33:03Z

I got an issue when I use serde before building window.
I don't know what to do with it : https://github.com/thiolliere/spring-crabs/

cactorium · 2017-05-03T11:45:47Z

What errors are you getting? It looks like it runs fine on my computer

gui1117 · 2017-05-03T17:15:18Z

it works on rustc 1.18.0-beta.1 (4dce67253 2017-04-25) but on rustc 1.17.0 (56124baa9 2017-04-24) it fails when running on the web:

Successfully compiled asm.js code (total compilation time 2664ms; storage initialization failed (consider filing a bug); 4 functions compiled slowly: $j:12:8 (507ms), fj:16:8 (324ms), Sj:13:8 (808ms), dj:6:8 (2602ms)) spring-crabs.js
emscripten_webgl_create_context failed: explicitSwapControl is not supported, please rebuild with -s OFFSCREENCANVAS_SUPPORT=1 to enable targeting the experimental OffscreenCanvas specification! spring-crabs.js:1:467981

thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: BackendCreationError(OsError("Error while calling emscripten_webgl_create_context: Internal error in the library (success detected as failure)"))', /checkout/src/libcore/result.rs:859 index.htm:141:13

note: Run with `RUST_BACKTRACE=1` for a backtrace. index.htm:141:13

gui1117 · 2017-05-03T20:40:44Z

I rewrite the example. Now it fails also on beta also.

Also I don't know if this is the right place for discussing on this issue. I think it is related to emscripten backend itself.

gui1117 · 2017-05-08T20:29:58Z

I have an issue related to hidpi_factor.

On some smartphone (I suspect with HDPI of 2) my game was only showing the bottom-left quarter of the expected view. I solved it by forcing hdpi_factor method to always return 1.

Also I found that our implementation of get_inner_size is not coherent with winit implementation.

I don't have any hardware with HDPI different from one to test.
I can propose to write get_inner_size_ as winit do and forcing hdpi_factor to one but that doesn't sound very good...

cactorium · 2017-05-10T04:01:08Z

Would something like this fix it? You probably have more experience with dealing with emscripten than me at this point, but it looks like using emscripten_request_fullscreen_strategy with scalingMode set to one of the EMSCRIPTEN_FULLSCREEN_CANVAS_SCALE_* options might be an easier (but maybe not better? I'm not sure) workaround

     #[inline]
     pub fn get_inner_size_pixels(&self) -> Option<(u32, u32)> {
-        self.get_inner_size()
+        self.get_inner_size().map(|(w, h)| {
+            let hdpi = self.hdpi_factor();
+            ((hdpi*(w as f32)) as u32, (hdpi*(h as f32)) as u32)
+        })
     }

Also what versions of stuff are you running? I can't build against the latest of everything because of the new EventLoop system (btw @tomaka would it make sense to make another PR to get this stuff working with the 0.8.0 version of glutin/newer version of winit?)

LukasKalbertodt · 2017-06-28T15:17:03Z

Any updates on this?

cactorium · 2017-06-28T23:56:02Z

Not as far as I know, but the code here's dated now, since winit switched the EventLoop thing at 0.6. I'm working on redoing the work here to the latest glutin, but there hasn't been any word from the maintainers about it. My fork's master branch might be usable right now, in case anyone needs emscripten working right now

aep · 2017-07-04T14:52:02Z

@cactorium actually your fork doesnt work out of the box, or am i missing something?

aep@derp: ~/kram/glutin cargo build --target=wasm32-unknown-emscripten --release
   Compiling lazy_static v0.2.8
   Compiling khronos_api v1.0.1
   Compiling bitflags v0.9.1
   Compiling libc v0.2.24
   Compiling log v0.3.8
   Compiling xml-rs v0.6.0
   Compiling shared_library v0.1.6
   Compiling winit v0.6.4
error[E0432]: unresolved import `self::platform::*`
 --> /home/aep/.cargo/registry/src/gh.neting.cc-1ecc6299db9ec823/winit-0.6.4/src/platform/mod.rs:1:9
  |
1 | pub use self::platform::*;
  |         ^^^^^^^^^^^^^^^^^^ Could not find `platform` in `self`

error[E0432]: unresolved import `this_platform_is_not_supported`
  --> /home/aep/.cargo/registry/src/gh.neting.cc-1ecc6299db9ec823/winit-0.6.4/src/platform/mod.rs:22:5
   |
22 | use this_platform_is_not_supported;
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ no `this_platform_is_not_supported` in the root

error[E0433]: failed to resolve. Could not find `Window2` in `platform`
   --> /home/aep/.cargo/registry/src/gh.neting.cc-1ecc6299db9ec823/winit-0.6.4/src/window.rs:113:22
    |
113 |         let w = try!(platform::Window2::new(events_loop.events_loop.clone(), &self.window, &self.platform_specific));
    |                      ^^^^^^^^^^^^^^^^^^^^^^ Could not find `Window2` in `platform`

[...]

cactorium · 2017-07-04T17:57:31Z

@aep Hmm, it looks like one of the pull requests I merged broke some stuff, and I can't be bothered with trying to fix it if it'll basically be stuck at the version 0.7 anyways. It looks like "emscripten_events" still builds correctly, so I guess you could use that one for now

francesca64 · 2018-07-09T16:16:33Z

I'm guessing this PR has been subsumed by subsequent additions to winit?

goddessfreya · 2019-02-26T00:42:44Z

Closing for inactivity.

cactorium added 2 commits February 12, 2017 14:35

Buildable for emscripten

191cd57

Minor cleanup

d9dac49

Remove duplication of winit from emscripten platform

3af812b

cactorium mentioned this pull request Feb 13, 2017

Add null platform for emscripten rust-windowing/winit#137

Closed

tomaka suggested changes Feb 13, 2017

View reviewed changes

Remove winit_window from emscripten target

a6c7877

tomaka approved these changes Feb 13, 2017

View reviewed changes

cactorium added 2 commits February 15, 2017 00:04

Add mouse callbacks

dfa621d

Add keyboard callbacks

aa471ad

cactorium added 2 commits February 15, 2017 09:31

Update key to virtual code mapping

600e1a6

Remove excessive logging

af606e2

Update winit version to 0.5.11 and remove unnecssary use statements

8ca74e5

Use canvas coordinates instead of client coordinates

7c1c4b9

thiolliere added 4 commits March 19, 2017 18:12

implement hidpi factor

acd4405

impl get set size

10b3b77

impl fullscreen and window dimensions

b8cdab7

cactorium mentioned this pull request Mar 20, 2017

Platform support for emscripten #856

Closed

cactorium and others added 8 commits March 20, 2017 21:12

Merge pull request #1 from thiolliere/master

4cb2ea0

more implementations

implement of touchevent

35322ce

fix mouse grab

9e3fab2

two fixes: * when mouse is grabbed send relative mousemove event * when mouse grabbed is refused, ask again.

catch more error

352e3ee

coding style

51ce5a3

consume event for pointerlock and fullscreen change

791260b

check result of get pointer lock status

f2f099c

Merge pull request #2 from thiolliere/master

21af3ab

Implement touch + fix mouse grab + catch more errors

Merge branch 'master' into master

599ef9f

gui1117 mentioned this pull request Sep 13, 2017

Emscripten rework rust-windowing/winit#286

Merged

iceiix mentioned this pull request Jan 12, 2019

Emscripten support roadmap #27

Closed

2 tasks

goddessfreya added the Closed-For-Inactivity label Feb 25, 2019

goddessfreya closed this Feb 26, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improve Emscripten support #857

Improve Emscripten support #857

cactorium commented Feb 12, 2017

tomaka commented Feb 13, 2017

cactorium commented Feb 13, 2017

tomaka commented Feb 13, 2017

cactorium commented Feb 13, 2017 •

edited

Loading

tomaka commented Feb 13, 2017

cactorium commented Feb 13, 2017

cactorium commented Feb 13, 2017

tomaka commented Feb 13, 2017 •

edited

Loading

tomaka Feb 13, 2017

tomaka Feb 13, 2017

cactorium commented Feb 13, 2017

cactorium commented Feb 14, 2017

tomaka commented Feb 14, 2017

cactorium commented Feb 15, 2017

cactorium commented Feb 15, 2017

gui1117 commented Feb 15, 2017

tomaka commented Feb 15, 2017

cactorium commented Feb 16, 2017 via email

cactorium commented Feb 17, 2017

gui1117 commented May 3, 2017 •

edited

Loading

cactorium commented May 3, 2017

gui1117 commented May 3, 2017 •

edited

Loading

gui1117 commented May 3, 2017

gui1117 commented May 8, 2017 •

edited

Loading

cactorium commented May 10, 2017 •

edited

Loading

LukasKalbertodt commented Jun 28, 2017

cactorium commented Jun 28, 2017

aep commented Jul 4, 2017 •

edited

Loading

cactorium commented Jul 4, 2017

francesca64 commented Jul 9, 2018

goddessfreya commented Feb 26, 2019

Improve Emscripten support #857

Improve Emscripten support #857

Conversation

cactorium commented Feb 12, 2017

tomaka commented Feb 13, 2017

cactorium commented Feb 13, 2017

tomaka commented Feb 13, 2017

cactorium commented Feb 13, 2017 • edited Loading

tomaka commented Feb 13, 2017

cactorium commented Feb 13, 2017

cactorium commented Feb 13, 2017

tomaka commented Feb 13, 2017 • edited Loading

tomaka Feb 13, 2017

Choose a reason for hiding this comment

tomaka Feb 13, 2017

Choose a reason for hiding this comment

cactorium commented Feb 13, 2017

cactorium commented Feb 14, 2017

tomaka commented Feb 14, 2017

cactorium commented Feb 15, 2017

cactorium commented Feb 15, 2017

gui1117 commented Feb 15, 2017

tomaka commented Feb 15, 2017

cactorium commented Feb 16, 2017 via email

cactorium commented Feb 17, 2017

gui1117 commented May 3, 2017 • edited Loading

cactorium commented May 3, 2017

gui1117 commented May 3, 2017 • edited Loading

gui1117 commented May 3, 2017

gui1117 commented May 8, 2017 • edited Loading

cactorium commented May 10, 2017 • edited Loading

LukasKalbertodt commented Jun 28, 2017

cactorium commented Jun 28, 2017

aep commented Jul 4, 2017 • edited Loading

cactorium commented Jul 4, 2017

francesca64 commented Jul 9, 2018

goddessfreya commented Feb 26, 2019

cactorium commented Feb 13, 2017 •

edited

Loading

tomaka commented Feb 13, 2017 •

edited

Loading

gui1117 commented May 3, 2017 •

edited

Loading

gui1117 commented May 3, 2017 •

edited

Loading

gui1117 commented May 8, 2017 •

edited

Loading

cactorium commented May 10, 2017 •

edited

Loading

aep commented Jul 4, 2017 •

edited

Loading