Bug in showLabel method - creates duplicate aria-describedby elements

[rodgerramjet]

Found in 1.13.0

This code (I assume) is trying to send the \b to the regex, but its being escaped by the JS

Should be something like "\b" + errorID + "\b"?

CODE:
} else if ( !describedBy.match( new RegExp( "\b" + errorID + "\b" ) ) ) {
// Add to end of list if not already present
describedBy += " " + errorID;
}

This creates loads of duplicate aria-describedby in the markup

[staabm]

//cc @tractorcow

[tractorcow]

My apologies for this error. While this is a bug, it didn't appear in testing due to the incorrect code being bypassed if showLabel could detect an existing error label to place this element into. When I wrote a few test cases I found this issue only could be replicated if I placed the aria-describedby reference prior to validation being setup. I hope the test case correctly captures your scenario.

[staabm]

@davidhedges could you verify that @tractorcow's PR fixes your issue?

[rodgerramjet]

Yep that fixes it- thanks guys :)

In terms of how I reproduced it- I haven't gone into too much detail - we are using jquery validate, unobtrusive validation, so a lot of stuff is getting done behind the scenes.

Here is a JSFiddle I threw together for you. Not sure if its an unobtrusive bug exacerbating your bug or not?

http://jsfiddle.net/cW9n5/

[rodgerramjet]

perhaps your test case should be updated to this scenario?

[rodgerramjet]

Updated Jsfiddle to show aria-described by on the page- test it by flling in invalid values & tabbing in and out of the field
http://jsfiddle.net/cW9n5/1/

[tractorcow]

It looks like this example breaks the code because there is no error label, so showError continually attempts to re-place it on the page. Is this the behaviour that unobtrusive validation circumvents?

[tractorcow]

I'm not sure we could update this test case to depend on third party libraries, if it is the case that this is necessary to replicate the full error. jquery.validate.unobtrusive.js isn't a supported module is it? The new test case I added failed under the old code, so that's probably enough (as far as our responsibility goes here) to validate we've fixed the issue.