Yes, I was also not convinced, that is why I tested for multiple
iterations. It seems to me the webrev.00 is the better "fix" as I saw
only 2-3 pixels differ which cause the exact match to fail. But,I am not
sure why in linux, the bufferedimages are exact. Maybe, there are some
differences with robot screencapture between platforms. Are you aware of
any?
Regards
Prasanta
On 3/1/2018 4:18 AM, Sergey Bylokhov wrote:
Hi, Prasanta.
It looks strange that half a second is not enough to hide a single
Swing component, isn't it?
On 28/02/2018 01:15, Prasanta Sadhukhan wrote:
Hi All,
It seems the test fails intermittently on windows as I can see 1 or 2
failures out of 15 iterations citing either "failing to hide opaque
button" or "failing to hide non-opaque button".
It did not fail for me in linux and mac in 15 iterations.
I have increased the delay before the robot screen capture when the
button is being hidden which causes it to pass on windows 7 on all 15
iterations. It does not affect ubuntu 17.10 and mac 10.13.3 where it
still passes.
webrev: cr.openjdk.java.net/~psadhukhan/7190978/webrev.01/
Regards
Prasanta
On 2/22/2018 9:51 PM, Prasanta Sadhukhan wrote:
Hi All,
Please review a testbug fix for an issue whereby it seen that it
fails citing "failing to hide opaque button".
It is seen that the test compares the buffered image strictly
pixel-by-pixel and even if it differs by 1,2 pixels it fails which
is what was happening here.
Proposed fix is to make it more lenient by allowing a tolerance of
20pixels difference in the comparison of 2 bufferedimage.
It is verified on windows.
Bug: https://bugs.openjdk.java.net/browse/JDK-7190978
webrev: http://cr.openjdk.java.net/~psadhukhan/7190978/webrev.00/
Regards
Prasanta
