fix unhandledrejection_event / onunhandledrejection edge compatibility

[dgoldstein0]

The current data says it's available in <= 79. this conflicts with caniuse (https://caniuse.com/?search=unhandledrejection) as well as not making much sense and also conflicting with my own testing - I verified it's available in Edge 88; and it's also been in Chrome for a while. As 79 is the first Edge based on the same code base as Chrome, it seems far more likely that Edge >=79 is the correct range of Edge versions with support - or certainly more correct than the <=79 range.

My test in edge was basically

window.addEventListener("unhandledrejection", (e) => console.log("unhandled rejection handler fired"));
Promise.reject("error")

The catch is that the rejected promise needs to be executed in the web page - if you reject a promise in the console, it doesn't trigger the unhandledrejection handler. The exception handler however can be set in the console.

and then saw the text console.log'ed

[foolip]

I can confirm this by testing Edge 18 and 80 with https://mdn-bcd-collector.appspot.com/tests/api/Window/onunhandledrejection. (I don't have access to Edge 79, but given this was added in Chrome 49, it's a very safe assumption to make.)

[foolip]

Thanks for contributing this fix @dgoldstein0, and welcome to BCD!

Just to explain what ≤79 even meant, it means that it was added in Edge 79, but we're not sure if it was also supported in some earlier version. So the previous data was effectively "don't know" for Edge 12-18, and "supported" for Edge 79 and later.

A lot of these "ranged versions" are the result of bulk conversions, and any that can be turned into a "real version" as you've done here would be great to fix. I think that if you search for "≤79" in particular you will find quite a lot of things that weren't in Edge 18, but verifying it and sending PR takes time. Thanks for doing so here!

[dgoldstein0]

ah, thanks for the explanation. I was definitely misinterpreting that to mean it wasn't supported after 79, which didn't seem right.

[foolip]

@dgoldstein0 yeah I can see how https://developer.mozilla.org/en-US/docs/Web/API/Window/unhandledrejection_event#browser_compatibility makes it seems like was only supported in 79 and earlier, which is the opposite of what it's intended to convey.

@ddbeck @Elchi3 FYI that there are other interpretations of the range values than intended. Perhaps you've seen this before, but I hadn't. It's a reasonable interpretation, not sure if there's a fix other than getting rid of as many range versions as possible?

[Elchi3]

@ddbeck @Elchi3 FYI that there are other interpretations of the range values than intended. Perhaps you've seen this before, but I hadn't. It's a reasonable interpretation, not sure if there's a fix other than getting rid of as many range versions as possible?

Unfortunately it came up a few times. Some improvements to the rendering on MDN were made, but yes, ideally, we only use ranged versions sparingly.

[ddbeck]

@ddbeck @Elchi3 FYI that there are other interpretations of the range values than intended. Perhaps you've seen this before, but I hadn't. It's a reasonable interpretation, not sure if there's a fix other than getting rid of as many range versions as possible?

Unfortunately it came up a few times. Some improvements to the rendering on MDN were made, but yes, ideally, we only use ranged versions sparingly.

Maybe we could revisit it again? Maybe transform ≤ into ? on MDN?

[foolip]

Since all of the allowed ranged versions are pretty far in the past now, I wonder if just treating them an exact version on MDN wouldn't work well in practice?

[Elchi3]

Since all of the allowed ranged versions are pretty far in the past now, I wonder if just treating them an exact version on MDN wouldn't work well in practice?

Can you open a new issue on this proposal and paste in a list of data points affected?

[foolip]

Sure, I've filed mdn/yari#3238.