Remove the dependency on the class variables _empty_item_index
and _cursor_part to make the code easier to follow. If
_update_completion is refactored in a similar way these variables can
be removed.
The test was flaky because waiting for scrolling didn't actually wait,
as the page logged a scroll position change to 0/0 directly after
loading.
We work around this by making the generic "And I wait until the scroll
position changed" not wait when it changed to 0/0.
With QtWebEngine, inserting text into the field is async, so if our
test runs too fast, it would fail. Now we instead log stuff via JS on
changes insteaad, and wait for those log messages in the tests.
We had some funny segfaults reported during scrolling (i.e. with
QApplication.sendEvent), and some code already had to use postpone=True
so there was no segfault...
So now we're back to postEvent again, and eliminated the main reason for
segfaults with it, which was re-using (and -posting) events which had
already been posted.
At least during tests this seems to run stable, let's hope it helps for
the people having crashes as well.
Turns out re.escape also escapes spaces, so we'd need to replace '(\\ )'
groups after escaping. At this point it's easier to just combine spaces
before escaping the pattern.
Fixes#1934.
Supersedes #1935.
On Windows, no echo.exe exists normally, so calling echo from the tests
is no good idea, since it relies on Cygwin to be installed and in %PATH%
(so that echo.exe is available).
This fixes this by providing a small echo.bat which is callable from the
tests, and then using a platform-specific path to the executable instead
of the hardcoded "echo". This should ensure that the tests pass even on
systems where echo.exe is not installed.
Note that we can't simply use a do-nothing exe (like rundll or hh.exe),
as we're passing parameters, and those executables may behave
differently in the presence of those parameters.
