On Tue, 27 Aug 2024 16:51:51 GMT, Phil Race <p...@openjdk.org> wrote:
>> Looks like it failed on Linux with the default pipeline, which should be
>> xrender. That's a bit surprising.
>> You should verify this and if necessary file a bug and exclude that failing
>> combination from testing.
>
> Also, I see you added metal.
> I think I'm going to request an extra layer on top of what I previously
> suggested.
>
> It is possible to have multiple @test tags, not just @run tags, so you can do
> platform-specific things on just the specific platform, so it would become
>
> /* @test
> @requires("os.name"="windows")
> @run main/othervm -Dsun.java2d.d3d=false
> @run main/othervm -Dsun.java2d.d3d=true
> */
>
> /* @test
> @requires("os.name"="macos")
> @run main/othervm -Dsun.java2d.d3d=false
> @run main/othervm -Dsun.java2d.metal=true
> */
>
>
> /* @test
> @requires("os.name"="linux")
> @run main/othervm -Dsun.java2d.xrender=false
> @run main/othervm -Dsun.java2d.xrender=true
> */
>
> and finally the one that is cross-platform AND defaults
> /* @test
> @run/main/othervm
> @run main/othervm -Dsun.java2d.opengl=false
> @run main/othervm -Dsun.java2d.opengl=true
> */
> Looks like it failed on Linux with the default pipeline, which should be
> xrender. That's a bit surprising. You should verify this and if necessary
> file a bug and exclude that failing combination from testing.
Yea, that's true, but it failed on a particular machine only and that too it
was a plain run without any flags(@run main/othervm
SwingButtonResizeTestWithOpenGL), the runs with xrender and other flags were
successful in the same machine. I just rerun it 10 times on different linux
machines and I couldn't observe the failure again(the latest results are
attached as a comment to the bug), so it looks like a machine specific issue
and I could see some UI debris from the previous runs on the screenshots.
anyway I will check it again with @test tags.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/20532#discussion_r1733264403
