Weekly tests with polyfill combinations are failing in IE8

see : https://github.com/Financial-Times/polyfill-library/runs/4280696106?check_suite_focus=true

Browser: ie/8.0 Test config: targeted / combined / shard 1   ✘ 610 tests, 46 failures  116 seconds to complete


Failures:
 - ie/8.0:
    -> TextEncoder passes the feature detect
       http://bs-local.com:9876/test?includePolyfills=yes&always=no&feature=TextEncoder
       false == true
 - ie/8.0:
    -> TextEncoder TextEncoder has a valid constructor
       http://bs-local.com:9876/test?includePolyfills=yes&always=no&feature=TextEncoder
       'TextEncoder' is undefined
 - ie/8.0:
    -> TextEncoder TextDecoder has a valid constructor
       http://bs-local.com:9876/test?includePolyfills=yes&always=no&feature=TextEncoder
       'TextDecoder' is undefined
 - ie/8.0:
    -> TextEncoder TextDecoder decodes encoded strings
       http://bs-local.com:9876/test?includePolyfills=yes&always=no&feature=TextEncoder
       'TextEncoder' is undefined
...

I can’t immediately see which change introduced this regression.

There is an exception on : polyfills/Object/defineProperty/polyfill.js:37 This explains why polyfill tests fail at some point. Something throws an exception, stopping execution of the polyfill bundle.

if (!supportsAccessors) {
	throw new TypeError(ERR_ACCESSORS_NOT_SUPPORTED);
}

But there are a lot of callers and most have a try catch.

To reproduce manually in IE8 :

http://bs-local.com:9876/test?includePolyfills=yes&always=no&feature=WeakMap&polyfillCombinations=yes

[JakeChampion]

I think I have now found the fix for this and also found another bug to fix 😄

[JakeChampion]

You’re right @romainmenke I did miss this - I’ll see if I have time to look into it today - thank you for pinging me on this!