On 5/31/19 7:09 PM, Sergey Bylokhov wrote:
Hello.
Please review the fix for JDK 13.
Bug: https://bugs.openjdk.java.net/browse/JDK-8214469
Fix: http://cr.openjdk.java.net/~serb/8214469/webrev.00
This test failure on macOS was reported as a possible regression of
JDK-6849922, but it is not strictly a regression.
Previously it is sometimes passed(because it checks the assertions too
fast), and now it always fails as expects because we now properly
check the assertion in the test.
The test has two components TextField and Choice. The Choice component
is non-focusable but the test opens the popup by the mouse and tries
to type something. On Windows, it is expected that the text will be
typed to the textfield, but on Linux it is expected that this typing
will be skipped(it was implemented this way to mimic the Motif
toolkit). On macOS the text will be typed to the textfield. I can say
that both behaviors are questionable but the difference in the
implementation is not strictly specified, and the test just defends
from the accidental change of these AWT components, so in the fix, I
just align expectation on Windows and macOS.
Taking into account the above, that states all 3 platforms have
different behavior, what makes you think that it is a test issue?
--Semyon
