According to issue 23979, this is still a problem in the Android 4.1.x native browser (Chrome's the default browser on this version, but the native one is often still installed).

It'd be good to capture this too. The check's getting a bit unreadable though! Maybe it's regex time?

if (/Android (2\.2|2\.3|4\.0|4\.1).* Mobile Safari/.test(ua) && ua.indexOf('Chrome') === -1) {
  return false;
}

Is there a way to programmatically detect a faulty History implementation without resorting to browser user agent sniffing?

@jokeyrhyme — Here's the original discussion on this: #733... general consensus was that it would be hard to detect (certainly smoothly), but very happy to try and improve this if we can think of a way.

This seems to cover versions which don't update location.pathname (e.g. Android 2.3):

Modernizr.addTest('history', function () {
  var oldLoc = window.location.toString();
  var result = false;

  if (window.history && 'replaceState' in window.history) {
    // Try changing history
    window.history.replaceState({}, '', oldLoc + 'a');
    result = window.location.toString() === oldLoc + 'a';

    // Put it back how we found it
    window.history.replaceState({}, '', oldLoc);
  }
  return result
});

But on 4.0.x and 4.1.x (maybe others too), the location object is updated correctly but the address bar isn't until refreshed. I have no idea how we can detect this without a UA sniff.

Interestingly, on 4.1.2 at least, the address bar is updated correctly if pushState() / replaceState() is called during page load — it works in a synchronous inline script, but doesn't on e.g. a click handler.

This snif has been broken for quite a while, any chance to get this merged? One should also add the sniff for Android 4.1 since it's still broken there. It works on Chrome on Android though, so it should be a bit more specific.
I don't like the test from @stucox, seems a bit risky to go modify the history, especially since we know it's buggy.

I can make a new PR with those changed ua sniffs, if it's gonna get merged :)

It has to be updated since it can't be merged as is. There are a couple other broken tests that need to be backported. Once this is updated, we'll probably ship another minor version

So should I make a new PR against master?

I can whip it up, but in general that is what needs to happen, yes.

The current thinking regarding user agent sniffing is that we should use feature detection to enable behaviours, and that it is acceptable to use user agent sniffing to disable behaviours. Obviously, we'd all love to live in a world where feature detection is all we need, but sometimes our only option is to explicitly blacklist specific browsers.

I just remembered, History.JS has special user agent detection in its blacklist as well:
https://github.com/browserstate/history.js/blob/0895348d1c8bfcbf6cda3d18ac607beded17bf57/scripts/uncompressed/history.js#L331
https://github.com/browserstate/history.js/blob/0895348d1c8bfcbf6cda3d18ac607beded17bf57/scripts/uncompressed/history.js#L305