On Mon, 25 Nov 2024 07:30:41 GMT, Alan Bateman <al...@openjdk.org> wrote:
> As part of JEP 491, and prep for (e.g. 8336254), we added more tests to > test/jdk/java/lang/Thread/virtual to stress several critical areas. A few of > these tests can take a long time on older macos-x64 hardware, esp. those with > HT enabled as jtreg -concurrency gets computed to a high value. Some of these > tests have spinning threads and end up competing with other tests. This > directory used to have a TEST.properties with exclusive.dirs=. to prevent > these tests running concurrently with each other but we decided to remove it > when the number of tests increased (at one point, tests were waiting several > minutes to execute because they couldn't run concurrently with other tests in > the directory). > > This PR proposes a few adjustments: > > - Dial down 3 tests (GetStackTraceALotWhenBlocking, > GetStackTraceALotWhenPinned.java, and ParkALot) on macosx-x64. This is done > in the test main to avoid doubling the number of test descriptions on these > tests. > - Dial down 2 runs of one test (TimedWaitALot #timeout-notify > #timeout-interrupt) on all platforms. No change to the other 2 runs of this > test. > - Add progress output/tracing to ParkALot. > - Remove "/othervm" from a few tests where that don't need it. These test-only changes look OK to me. Is it intentional that only one of the test definitions of `SleepALot` has been updated to remove `/othervm` instead of both? ------------- Marked as reviewed by jpai (Reviewer). PR Review: https://git.openjdk.org/jdk/pull/22357#pullrequestreview-2458473660
