I have looked at the embind issue with async wasm compilation.
I was able to identify three (mostly ?) separate issues here :
- First, compiling with --bind produces code that doesn't initialize
properly. And that, even if the code does not use embind at all.
So, normal code compiled with --bind is not working.
What happens is that a big portion of code is generated differently when
specifying --bind. It does not happen with only --std=c++11, though, so I
don't really know how embind can change the compilation output that much.
To give an example of a change that breaks initialization :
var _main = Module["_main"] = function() { return
Module["asm"]["_main"].apply(null, arguments) };
becomes :
var _main = Module["_main"] = asm["_main"];
To fix that in a compiled code, I simply copied the worling version of
the code portion and added the lines for the missing symbols.
Once this is fixed, code compiled with --bind not using embind will
initialize and run normally (at least the hello world I tested).
- Second issue is with initialization of code that uses embind.
The error given is "No dynCall invoker for signature..."
When looking for an invoker in requireFunction, embind uses the asm
global object.
Replacing asm with Module["asm"] fixed the issue.
After that, code using embind initializes correctly.
- However (and this is the third issue), functions called through embind
do nothing. I am still investigating this one, maybe it is just due to the
runtime exiting.
For now, I don't know how to fix the first issue for good and I am not
entirely sure about my fix for the second one.
Will keep you updated.
Le jeudi 23 février 2017 22:21:57 UTC+1, Alon Zakai a écrit :
>
> Hi everyone,
>
> Embind is a tool emscripten has that helps connect C++ and JS,
>
>
> https://kripken.github.io/emscripten-site/docs/porting/connecting_cpp_and_javascript/embind.html
>
> The original creators have been less active lately, and it seems that
> currently it is no longer actively maintained. In particular, issues get
> filed about embind that sometimes get no response.
>
> On the positive side, that issues get filed means this is a feature that
> people want. Also, we do have good testing for embind and what is tested
> should work perfectly. And while there are other mechanisms for connecting
> C++ and JS - ccall/cwrap for simple calls, the WebIDL binder for binding
> C++ to JS - embind does a bunch of things that have no alternative
> currently.
>
> So embind is definitely a useful and usable feature now, however, without
> filed issues being triaged and fixed, and no new development, that might
> change eventually. For example, a specific issue I noticed over the last
> few days was that async wasm compilation - a new feature - doesn't work in
> embind, we need help with things like that.
>
> The bottom line is that for embind to keep up with emscripten development,
> it needs someone to work on it. Is someone interested to help out?
>
> - Alon
>
--
You received this message because you are subscribed to the Google Groups
"emscripten-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
