Win 7 with IE11 (using ModerIE) selenium-webdriver 3.3.0 IEDriverServer_beta_Win32_3.3.99.1
I can confirm that I can reproduce SeleniumHQ/selenium#3587 using WatirSpec. It fails using IEDriverServer 3.3, but does not failing using beta, so it was somehow fixed.
I have only one spec that fails on beta but passes on 3.3. The specs are about drag and drop functionality. While basic "drag element and drop to another element" passes, there are few that are failing:
- Drag element that is out of viewport.
- Drop on element that is out of viewport.
- Drag element and drop by offset.
Ruby bindings don't work with W3C dialect of IEDriverServer, so I needed to hack it around in order to make specs run, but I still need time to turn these failures into bug reports. I hope to do that during the weekend.
Thanks very much for doing this, it's much appreciated. It's gratifying that so much of it Just Works™.
The window switching spec failures are protocol differences in that a different end point is used for the W3C protocol dialect for window size and position manipulation (
GET|POST /session/{sessionId}/window/rect
instead of separate end points for size and position). This will need to change in Ruby because the next geckodriver will also implement the proper end points here.The drag-and-drop spec differences are interesting. They're unexpected, but I also wonder if they're not obsolete for IE. What happens with those specs for geckodriver? I suspect they either throw an expected exception ("move target out of bounds"), or they're guarded against on that browser because they would throw that exception. The spec is very clear that attempting to move the mouse outside the view port isn't supported, even in the drag-and-drop case. The beta driver doesn't explicitly fail in that way, and those failures are unexpected, but there may be some side effect of the refactoring that's causing it to work in a spec-compliant way anyhow.
The difference in radio buttons is concerning. I'll need to investigate that one a little further.