On Mon, 4 Nov 2024 04:15:09 GMT, David Holmes <dhol...@openjdk.org> wrote:
>> Hi all, >> Sometimes some tests only fails or crash run by fastdebug/slowdebug jdk >> binary, such as >> `java/lang/Thread/jni/AttachCurrentThread/AttachTest.java#id1` which has >> been recorded by [JDK-8343244](https://bugs.openjdk.org/browse/JDK-8343244). >> To make less CI noisy, we can disable these tests by add tag such as >> `@requires vm.debug == false`, but I think put the releated tests to >> Problemlist is a better way before the root cause failure has been fixed. >> Thus I think it's necessary support Problemlist for fastdebug/slowdebug >> build, the Problemlist only work for fastdebug/slowdebug build. >> By the way, the timeout factor with fastdebug/slowdebug should larger than >> release build, this PR also make jtreg timeoutFactor to 10 if the tested jdk >> is fastdebug/slowdebug build. > > make/RunTests.gmk line 860: > >> 858: >> 859: ifneq ($$(DEBUG_LEVEL), release) >> 860: JTREG_AUTO_PROBLEM_LISTS += ProblemList-debug.txt > > We (Oracle) typically do not need this because of the way we run our tests. > Hotspot tests are always run on fastdebug in lower tiers and only some tests > run on release bits in higher tiers. So the fact the ProblemList affects > fastdebug and release is not really an issue. Conversely, JDK tests are run > on release bits in lower tiers, and only in some cases run on fastdebug in > later tiers. > > My main concern here is that we end up with too many potential ProblemList > files that have to be managed and maintained. And, IMO jtreg should be > enhanced so that the PL syntax allows more flexibility over the conditions > under which a test is to be excluded. Thanks for your detail explanation, your concern about too many potential Problemlist files to maintained is reasonable. If no one else support this change, I wll close this PR and the releated issue as `Won't Fix` ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/21862#discussion_r1827253052
