Aborted requests on IE9 are not completed

[montgomery1944]

In one of the applications we had to write a mechanism of aborting previous requests, when the new will be fired, you can see the simplified version here: http://plnkr.co/edit/ntDsVwDgG8digzjxBfKX.

The problem is, that after the change introduced in #9329 abortion of requests on IE9 stopped working, because "onabort" event is not supported there, which means that "completeRequest" function won't be called at all on abort.

In provided plunker you will see alert with message "ABORT" on all supported browsers except IE9.

Please have on mind, that IE9 is behaving weirdly when aborting the request, which was already fixed in Angular in #4940.

[realityking]

Rats. I don't have time to look at it right now as we're having a big release next week, after that I'll try to take a look at it if nobody beats me to it.

A cursory Google search shows that there seem to be quite a few issues with aborting request in IE9 where the request isn't really aborted. However I don't see how #4940 relates to it as we're not accessing the xhr object in the onabort handler.

[pkozlowski-opensource]

@realityking I agree that it doesn't seem to be related to #4940.

Then, when it comes to aborting request on IE - it would be awesome if you could have a look at this one! I'm moving it to 1.3.6 so this issue doesn't get forgotten.

[petebacondarwin]

@realityking - did you get time to take a look at this problem?

[realityking]

Sorry, took me a while to get to this. First the plunkr can not be used directly as IE9 does not support CORS. I created a slightly different one to test: http://plnkr.co/edit/M1NbCR752zhHqYaWfl2D

The problem is however relatively simple, IE9 just doesn't implement onabort at all. The following events are implemented:

• onload
• ontimeout
• onreadystatechange

Note that onerror is also missing.

This leaves us in a pretty bad spot. If we don't want to revert to onreadystatechange (with all the issues that it had) the only thing we could do is manually call completeRequest when the browser is IE9. I'll create a pull request for that.

[laurelnaiad]

@realityking, is there a PR for this that I could turn into a monkey-patch for my project? As you can see I've sort of worked around the problems this is causing in e2e tests by crashing the tests more quickly so they don't stall, but the app itself is hanging at a spinner due to never getting a timeout from the request when travis is slow. I'm sure I could cook up something ugly in an interceptor, but if you have something lying around, it might be helpful. :)

[lorenzodallavecchia]

In addition to IE9, some Mobile Safari versions seem to be affected.

This problem also happens when the request times out by itself (with the default browser timeout), because in that case ontimeout is called, and Angular is not listening to that one.
Fortunately, that issue can be mitigated by specifying a timeout when calling $http (Angular will complete the request when the time runs out).