On Thu, 26 Mar 2026 13:14:40 GMT, Chen Liang <[email protected]> wrote:
>> When running on the windows debug builds, this test displays greatly varying >> timeouts. In one failing run, the time taken ranges in iterations are >> 58.388, 22.764, and 63.717 seconds. Another run has 8.814, 38.341, 25.260, >> 14.678, and 105.368 seconds. I think we can probably adjust this await time >> by the timeout factor to be more permissive for now. >> >> @JornVernee Would you be able to diagnose the cause of such perturbations in >> closing times and find a better solution? Could this be a CI issue? > > Chen Liang has updated the pull request incrementally with one additional > commit since the last revision: > > Switch to use close I submitted the updated version on CI against windows x64 fastdebug; 1 out of 64 runs used 44 secs but the other all finished in 20 secs. Same results for the baseline 64 runs, 1 run used 22 secs, all other in 20 secs. I double checked with @JornVernee, the old timeout is just a manual timeout mechanism that can be safely replaced by the JTReg one. I think we will go with this version (more leniency on one of the iterations) and re-investigate if windows scheduler proves itself to be problematic. ------------- PR Comment: https://git.openjdk.org/jdk/pull/30447#issuecomment-4134791916
