Add ArrayBuffer fallback for #7918

[bspeice]

What kind of change does this PR introduce?

Adds a failure case so Electron-style apps can easily use WebAssembly

Did you add tests for your changes?

Doesn't run a headless test, not sure how best to add test support (is tested using a local Electron app)

Does this PR introduce a breaking change?

No

What needs to be documented once your changes are merged?

Hopefully nothing? Think there may be a README example that needs updating, but this should be something that "just fixes" some background issues.

[jsf-clabot]

Thank you for your submission, we really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

[webpack-bot]

For maintainers only:

• This need to be documented (issue in webpack/webpack.js.org will be filed when merged)

[xtuc]

That's a great solution.

In the issue I thought you wanted to inline the binary, which I would be against to.

[webpack-bot]

Thank you for your pull request! The most important CI builds succeeded, we’ll review the pull request soon.

[bspeice]

Alright, continued working with this, and bad news - this patch works fine for --mode=development, but --mode=production has issues:

CompileError: WasmCompile: Wasm decoding failed: field name: no valid UTF-8 string @+1791
    at file:///home/bspeice/Development/isomorphic_rust/percy_patched_webpack/dist/bundle.js:1:2658
    at <anonymous>

Character 2658 is the actual call to WebAssembly.instantiate, and character 1791 doesn't seem to be relevant anywhere. Is this something that should still go through?

[xtuc]

@bspeice could you please try again with this patch #7732?

[bspeice]

Can confirm that #7732 does in fact fix the issues I saw. --mode=production runs fine on that version.

[bspeice]

@xtuc - Anything further needed on this? Everything seems to be working from my end, want to make sure I didn't miss something.

[xtuc]

Could you please add tests on that?

I guess you can overide the WebAssembly.instantiateStreaming function to make it throw and test the fallback?

[TheLarkInn]

@xtuc How's this looking. Passes tests.

[xtuc]

@TheLarkInn do we know if that's a common issue in pratice?

I'm not personally blocking it, but i'm not happy with the solution. There are many reasons for TypeErrors during instantiateStreaming.

However, the worst case would be to call instantiate twice. I'm pretty sure it won't run two compilations, most of the type mismatch are coming from the ImportObject, which is checked before.

Ideally, either Electron should figure out a way to correctly load Wasm or the mime-type restriction needs to change.

[bspeice]

Given that both Electron is unlikely to substantively change Blink to allow this, and the spec is unlikely to change, is this option acceptable? If not, I'd rather close out the request for now to clean things up.

[bspeice]

As noted in #7918, this is fixed simply by using Webpack 5 (and enabling the asyncWebAssembly experiment). This change is no longer necessary.