Any tests that rely on shared state that can’t be safely multithreaded needs to be categorized as such. I thought I covered everything in API already, but perhaps not.
On Wed, Mar 24, 2021 at 21:38 Ralph Goers <ralph.go...@dslextreme.com> wrote: > I believe Matt made the tests in the API run in parallel. I’ve often > wondered about that since LogManager is creating contexts meant to be > shared across multiple threads. I have no idea how all that will behave in > a multi-threaded environment since most unit tests create and destroy > LoggerContexts. > > Ralph > > > On Mar 24, 2021, at 4:25 PM, Remko Popma <remko.po...@gmail.com> wrote: > > > > Wild guess: multi-threading issue? (That's my go-to boogeyman) > > :-) > > > > On Thu, Mar 25, 2021 at 6:03 AM Carter Kozak <cko...@ckozak.net> wrote: > > > >> The flake appears to be resolved, just as mysteriously as it appeared. > >> > >> -Carter > >> > >> On Wed, Mar 24, 2021, at 10:28, Carter Kozak wrote: > >>> Any ideas why org.apache.logging.log4j.EventLoggerTest.structuredData > >> would fail due to the following? > >>>> java.lang.NullPointerException > >>>> at > >> > org.apache.logging.log4j.EventLoggerTest.structuredData(EventLoggerTest.java:42) > >>> > >>> I haven't been able to reproduce this failure locally, but it appears > to > >> have occurred both on release-2.x and a PR branch I pushed later. > >>> -Carter > >>> > >> > > >
