fix(core): panic on big string allocation

[bartlomieju]

Fixes #6649

[piscisaureus]

Hmmm

Good to see this fixed but IMO returning an empty string is a really weird response - why not throw a RangeTypeError?

I actually wonder if this might be a bug in Chrome rather than intentional behavior. I could see that happen because of a weird quirk in the v8 api, which is that when a V8 API function returns an empty MaybeLocal this always means a JS exception was thrown. With one exception: v8::String::new() can also return an empy MaybeLocal but in that case it just means the string was too long.

[piscisaureus]

Spec: https://encoding.spec.whatwg.org/#dom-textdecoder-decode

[bartlomieju]

Hmmm

Good to see this fixed but IMO returning an empty string is a really weird response - why not throw a RangeTypeError?

I actually wonder if this might be a bug in Chrome rather than intentional behavior. I could see that happen because of a weird quirk in the v8 api, which is that when a V8 API function returns an empty MaybeLocal this always means a JS exception was thrown. With one exception: v8::String::new() can also return an empy MaybeLocal but in that case it just means the string was too long.

Spec: https://encoding.spec.whatwg.org/#dom-textdecoder-decode

@piscisaureus thanks for pointing this, indeed spec suggests to use TypeError. I think it's better solution that silently ignoring the error and returning empty string. I've updated the PR.

[piscisaureus]

Had a change of heart, switched it to RangeError after all, and made some minor tweaks to the doc comment.
Now it looks good to me.