On 11/7/19 6:29 am, Philip Race wrote:
The fix seems fine but if the app isn't setting null how are we getting null
here ?
We get null here only if the test/app sets the null value, in which case this
null
value will be ignored and some other font will be used by the peer.
Frequent hierarchy validation does not in itself explain that to me.
On HiDPI systems the invisible components use some default fonts
render_context/metrics
and when the component became visible we change these default values to the
actual
HiDPI graphics config ->to a different fonts render_context/metrics -> need to
validate
hierarchy->NPE.
-phil.
On 11/6/19, 2:48 PM, Sergey Bylokhov wrote:
Hello.
Please review the fix for JDK 14.
Bug: https://bugs.openjdk.java.net/browse/JDK-8233657
Fix: http://cr.openjdk.java.net/~serb/8233657/webrev.00
During testsprint I have analyzed some intermittent failures of our tests and
found some intermittent NPE. The root cause is the calling of
Component.validate()
if the null font is used. The bug reproduced mostly on HiDPI systems where we
validate the hierarchy of components more often.
The validate method is aupdated to use the same logic as setFont(), where null
value
is ignored and never passed to the peer.
Actually the fix is simpler than the tests problem listing due to product bug.
--
Best regards, Sergey.
