[webkit-reviews] review requested: [Bug 39418] history.pushState doesn't work for the first page in a window. : [Attachment 57091] layout test v2

[bugzilla-daemon at webkit.org]

Jędrzej Nowacki <jedrzej.nowacki at nokia.com> has asked  for review:
Bug 39418: history.pushState doesn't work for the first page in a window.
https://bugs.webkit.org/show_bug.cgi?id=39418

Attachment 57091: layout test v2
https://bugs.webkit.org/attachment.cgi?id=57091&action=review

------- Additional Comments from Jędrzej Nowacki <jedrzej.nowacki at nokia.com>
(In reply to comment #8)
> (From update of attachment 57020 [details])
> (...) But run-webkit-tests reuses the same WebView for a series of tests. 
Therefore - unless it goes first - this test won't actually excercise the buggy
code path.
Wow, I didn't know, it is evil :-).

(In reply to comment #7)
> If a layouttest was catching this case, then that layout test should've been
failing.
> If a layouttest was NOT catching this case, then adding one would not be
redundant.
I thought that bug was exposed after some other changes. For me
document-destroyed-navigate-back-with-fragment-scroll.html is failing without
the patch, but only when it is run as a first test, I haven't noticed ordering
issue. It is why I wrote that a new test would be a bit redundant.

(In reply to comment #9)
> (...) There is a way to get DRT to run the test in a fresh WebView *even* in
the context of run-webkit-tests, and that is to run the test in a new window.
New patch use this trick. Shouldn't we fix DRT? The bug wasn't caught only
because of this "feature", for example
document-destroyed-navigate-back-with-fragment-scroll.html covers similar use
case.