RFR: 8275170: Some jtreg sound tests should be marked headful

[Sergey Bylokhov]

On Sun, 24 Oct 2021 16:45:15 GMT, Phil Race <prr at openjdk.org> wrote:

> This makes the tests eligible to be run on headful systems. In practice it seems like ONLY headful systems actually have the sound devices. The sound keyword isn't understood by anything yet, so adding headful is the only way we have right now of ensuring these tests are eligible to be run on systems that have the necessary support.

This is not the right approach, it was made a hard work to eliminate the headful keyword from any tests including sound which are not throw the headless exception or something. The "headful" means that the test will fail otherwise due to UI session missing. We should not abuse it to select some other unrelated tests. Like we do not doing this for printer.

The problem you try to solve is that you have a "bad" system where nothing is configured and the "good" system where UI/sound(what about printer?) are configured. And you want to exclude all tests from the "bad" host and run them on the "good" one by using the headful keyword which is not targeted for this usecase.

> As I explained in the initial PR description headful is also useful for implying exclusive access.

Such exclusive access should not be needed by the sound, in fact it was found a few hard too spot concurrency bugs when we start to run the sound tests in parallel.
 
> And without this keyword it means that these tests are ALWAYS run on systems without sound device support.

Is it always just because the sound is not configured on the systems where you run the tests without headful keyword? How it is related to the "headfulness" of the systems?

> So for years we've effectively problem listed these tests due to the tests not having headful and silently passing when there's no sound. Without headful no one runs these on systems that have the needed device.

This was done intentionally that everybody who run the tests also run the tests for the sound, moreover it was included the openjdk tier3, before it was accidentally removed from there. 

> The 2 keywords give flexibility - anyone who wants to filter when its marked headful and needing sound can do this, but just adding sound right now will solve nothing.

Such flexibility exists already, it is possible to run test/jdk/javax/sound on any computer you want and exclude it from the jdk_desktop test run.

> And there are a couple of tests NOT in OpenJDK that already do this so we have precedent - and in a previous life one was added by you and the other approved by you ..

If it is not in the OpenJDK then we do not need to discuss it here and that's not actually a precedent. We should investigate each case one by one, the headful keyword does not solve any issue on windows, since all windows are headful(or most of them where we interested in the audio), on macOS the sound subsystem is built-in. And on Linux it is possible to configure sound as well and it will work fine w/o X server.

I think most of the sound tests are written according to the spec which does not required UI, and if it was necessary to make it pass by running on the system which have UI system configured make me think that some product bug was workaround.

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

PR: https://git.openjdk.java.net/jdk/pull/6086