RFR: 8276058: Some swing test fails on specific CI macos system [v2]

[Phil Race]

On Fri, 29 Oct 2021 05:38:55 GMT, Sergey Bylokhov <serb at openjdk.org> wrote:

> > Broader ? Please expand since the change you suggest is only about the cursor
> 
> This issue JDK-8276058 is broader than the cursor, related, this is why I would like to exclude it from it.

Still not explained. Broader why ?
> 
> > The test part of the fix has multiple stability improvements, and there's no reason to throw that out. And moving the pixel capture position to a more safe position is fine too. I just think that -10 isn't sa\fe enough.
> 
> Sure, I have no comments about the changes related to stability, but there are suspicious cases like using tolerance.

meaning -10 .. yes .. we can and should tweak that and if there's still an issue deal with it as it arises
> 
> > Who mentioned aarch64 ? Are you referring to you proposed change. I don't see the aarch64 connection anywhere FWIW the actual hardware on which this was seen is (I think) a 2013 Mac Pro running 10.15.7
> 
> I have mentioned that a few comments above, on aarch64 the cursor is always present on the screenshot, unlike x64. both systems are "similar".

I missed that but these failures are NOT seen on aarch64. If it was a feature there we'd have seen it.
We've tested on M1 imac and M1 macmini -- no problem. 
> 
> BTW I have found that the built-in screencapture tool has the ability to take a screenshot with and w/o cursor, so probably there is a way to specify that as well via some public API. Or maybe it is so fast that the cursor flickering is invisible.

I pointed that out earlier .. even before you raised the cursor size issue.

-------------

PR: https://git.openjdk.java.net/jdk/pull/6140