IE < 10: console isn't always available, console methods are all functions.

[krusynth]

console isn't defined in IE 8 & 9, except when the Developer Tools are open (and they aren't on automated tests). Even when the Developer Tools are open, all of the methods on console are defined such that you can't call apply() on them (as we do in our compiled version of Util.deprecationWarning

console-shim looks promising, but really it's only a few lines to shim it.

Relates to #408

[tilgovi]

If you just require('console') anywhere we use it browserify will include https://www.npmjs.org/package/console-browserify with no extra effort, I think. We may not even need to add it to package.json because it's expected to be a global, included-by-default, node module.

[aron]

Or could we just do console && console.log()? Do we want developer messages enough in IE8/9 to include an entire shim?

[krusynth]

@tilgovi That package doesn't work for IE8.

[krusynth]

@aron That would be possible, there aren't that many places where it's happening.

[krusynth]

@aron Looking closer at this, there are about 30 instances where console is being used - including some sinon stubs in tests. I'm afraid that it'd be more than just checking for console existing, we'd need to wrap it as well - due to the way that IE8 defines it when it's present, it can't be .call'd or .apply'd. I've been using console-shim which has been working perfectly for everything for the last week or so.

With regard to:

Missing browser functionality (in IE8) that can be patched using polyfils should be included by the consumer and not packaged in the codebase (this includes including the shims in the IE8 test suite).

How do we want to actually handle this? Just leave a note saying that users need to add this to their package.json themselves? For the other packages I've just been using conditional tags in the runner, but console-shim is not available on cdnjs.

[aron]

How do we want to actually handle this? Just leave a note saying that users need to add this to their package.json themselves? For the other packages I've just been using conditional tags in the runner, but console-shim is not available on cdnjs.

Yeah, I think we should document the libraries we recommend, and examples of where to find them, but leave the implementation up to the user. If we have the resources, creating an IE polyfil repo with a built version of all the scripts would be great, then implementors can just include this within IE conditional comments.

[tilgovi]

I don't think having apply is important. I know even browsers other than IE I've seen that console functions aren't "normal" functions (try passing a reference to console.log around or binding it to a different this, and you'll see inconsistencies).

Have we any idea why console-browserify doesn't work in IE8? Whether @Raynos would accept a patch? Otherwise, using a different shim seems fine rather than using require('console').

[krusynth]

We actually use apply in places like this one, but also in how coffeescript translates things like this into :

return console.warn.apply(console, ["Annotator DeprecationWarning:"].concat(__slice.call(args)));

I'm not sure what doesn't work for console-browserify - but frankly I'd rather not spend the time trying to debug and support yet another project when there's an existing fix that works. I'm going to go ahead and commit console-shim to my grunt-test branch as a dev dependency along with the others - and I still need to add to the documentation for all these IE8-only notes.

[tilgovi]

but frankly I'd rather not spend the time trying to debug and support yet another project when there's an existing fix that works

Except that require('console') is quite nice and explicit and browserify handles the shim automatically. Here's the link if anyone wants to look: https://ci.testling.com/Raynos/console-browserify

[krusynth]

No argument here that that's a preferable implementation - but a working solution is better than a not-working solution at this point.

[tilgovi]

I know the testling results say IE8 fails, but I've just run the tests for console-browserify in IE8 in a VM from modern.ie and it works fine.

[tilgovi]

That means that unless I'm missing something, require('console') and dropping our own console shim is the way to go.