Fix IE10 crypto

[funkyboy]

In particular this makes the following tests pass also on IE10:

2.generateRandomKey0 (2, 0, 2)
3.generateRandomKey1 (2, 0, 2)
4.getDefaultParams_wordArray_key (2, 0, 2)
5.getDefaultParams_base64_key (2, 0, 2)
6.getDefaultParams_check_keylength (2, 0, 2)
7.getDefaultParams_preserves_custom_algorithms (2, 0, 2)
15.single_send_128_with_web_socket_binary_transport (2, 0, 2)
16.single_send_128_with_web_socket_text_transport (2, 0, 2)
17.single_send_128_with_xhr_streaming_binary_transport (2, 0, 2)
18.single_send_128_with_xhr_streaming_text_transport (2, 0, 2)
19.single_send_128_with_xhr_polling_binary_transport (2, 0, 2)
20.single_send_128_with_xhr_polling_text_transport (2, 0, 2)
21.single_send_128_with_binary_transport (2, 0, 2)
22.single_send_128_with_text_transport (2, 0, 2)
23.single_send_256_with_web_socket_binary_transport (2, 0, 2)
24.single_send_256_with_web_socket_text_transport (2, 0, 2)
25.single_send_256_with_xhr_streaming_binary_transport (2, 0, 2)
26.single_send_256_with_xhr_streaming_text_transport (2, 0, 2)
27.single_send_256_with_xhr_polling_binary_transport (2, 0, 2)
28.single_send_256_with_xhr_polling_text_transport (2, 0, 2)
29.single_send_256_with_binary_transport (2, 0, 2)
30.single_send_256_with_text_transport (2, 0, 2)
31.multiple_send_binary_2_200 (2, 0, 2)
32.multiple_send_text_2_200 (2, 0, 2)
33.multiple_send_binary_20_100 (2, 0, 2)

[funkyboy]

[funkyboy]

Also IE9. I am surprised!

[SimonWoolf]

This is the sort of check that needs to be encapsulated in the Platform object. As it stands, the effect of this change will be to force weak random numbers for nativescript, react-native, and IE11 (which uses window.mscrypto rather than window.crypto), all of which do support strong random numbers via Platform.getRandomValues. (Or just crash in the case of platforms with no window object!)

To put it another way, the code already checks that Platform.getRandomValues is defined; the bug is that it shouldn't be defined if in a browser context where you have neither window.crypto nor window.mscrypto.

[SimonWoolf]

If the parsing issue is just that they don't like the newline before the method call with the dot syntax, would be clearer to just move it up to the previous line, avoid having to remember the rules of ASI

[funkyboy]

I tried using

}).then(function (results) {
	callback(null, results.content.toString());
}).catch(function(err) {
	callback(err);
});

but no luck :(

[funkyboy]

@SimonWoolf Is this preferable?

}).then(function (results) {
	callback(null, results.content.toString());
})['catch'](function(err) {
	callback(err);
});

[SimonWoolf]

ok, so they're confused by catch being a keyword. Yup, I'd prefer that latter one to the original, as then it makes it clearer that it's a property access rather than an array literal, thanks

[funkyboy]

@SimonWoolf did you mean something like this?

[SimonWoolf]

Sure, that should work. can you confirm it works as expected in IE10 etc.?

[funkyboy]

@SimonWoolf Yup.
Tests keep on passing on Firefox and Chrome.
And with this also on IE10.
Attaching some screenshots for reference.

chrome.pdf

ff.pdf

[SimonWoolf]

Thanks @funkyboy