Hi Sergey,
The probability of such scenario is very low but I agree it doesn't
equal to zero.
Also, there is yet another scenario: right after the request to open
candidate window is scheduled in the native queue the parent window is
closed.
The updated fix: http://cr.openjdk.java.net/~ssadetsky/8179665/webrev.01/
--Semyon
On 05/10/2017 03:48 PM, Sergey Bylokhov wrote:
Hi, Semyon.
It seems that this new check is executed on a toolkit thread, and does not
solve the problem if the window became invisible after this check but before we
post an event to the EDT.
I guess that the root cause is that getTextLocation() throws an unexpected
error but it should handle it and return empty rectangle(I guess it can be
tested just by calling this method when the component is invisible). The
current test and a check is also useful.
----- semyon.sadet...@oracle.com wrote:
Hello,
Please review fix for JDK10:
bug: https://bugs.openjdk.java.net/browse/JDK-8179665
webrev: http://cr.openjdk.java.net/~ssadetsky/8179665/webrev.00/
The root cause of the issue is IME events are sent when always-on-top
state is set or cleared even during the window is not visible. In this
case requesting the window position causes the exception.
The fix introduces a check for window visibility before the
calculation
of the IME status window position. The latter should not be shown when
the parent window is not visible so the IME window position is
unnecessary.
--Semyon
