I'd prefer not to remove JDK 8 support. Java 8 is still widely supported and used.
https://en.wikipedia.org/wiki/Java_version_history It simplifies our build a bit but I think we'll end up getting railroaded into supporting 2 branches - one that supports that Java 8 and gets critical patches and another that supports Java 11 and gets all the fixes. On Sunday 14 May 2023 at 15:36:17 IST, Dominik Stadler <dominik.stad...@gmx.at> wrote: Hi, please note that setting the property likely has no effect with the additional jar-file and options. You need to download the corresponding jar-file and add the -Xbootclasspath option to actually switch rendering, see e.g. https://github.com/centic9/poi-benchmark/blob/master/src/jmh/java/org/apache/poi/benchmark/suite/BaseBenchmark.java#LL143C1-L143C1 But simply starting to deprecate JDK 8 for the next version is fine with me as setting this up properly proved to be very error-prone because it is not failing, but silently ignores missing stuff and silently falls back to the broken renderer. Regards... Dominik. On Sun, May 14, 2023 at 12:58 PM PJ Fanning <fannin...@apache.org> wrote: > Thanks Dominik. I changed the poi-ooxml gradle file for JDK8 to use > MarlinRenderingEngine in tests. I think that helped a little. The builds > seemed to hang in random places before. > > The remaining issue appears to be the poi-integration tests. I disabled > them for JDK8 (again in the gradle file for that subdir). The CI builds are > now clean. > > If someone wants to experiment and find which file(s) is causing the > poi-integration JDK8 test issues, feel free to do so. The key thing is that > poi-integration tests run ok on newer JDKs so we have coverage. > > On 2023/05/09 18:24:54 Dominik Stadler wrote: > > Hi, > > > > This might be caused by "busy loop hangs" in the Java > > "PiscesRenderingEngine", I had a similar issue in poi-benchmark > > <https://github.com/centic9/poi-benchmark> , I solved it by switching to > > the "MarlinRenderingEngine" as described in the second "Note" at > > https://poi.apache.org/components/slideshow/ppt-wmf-emf-renderer.html > > > > I download the additional jar at > > https://github.com/centic9/poi-benchmark/blob/master/runBenchmark.sh#L13 > > and patch the build.gradle file at > > > https://github.com/centic9/poi-benchmark/blob/master/src/jmh/java/org/apache/poi/benchmark/suite/BaseBenchmark.java#L144 > > > > Not sure what actually triggers the endless-loop, maybe a specific > > patchlevel of the JDK or maybe some timing needs to fall in place. > > > > It is really annoying that the JDK 8 does not seem to get a fix for this > > any more. > > > > We could try to make this a "permanent" in our tests, but it would add > even > > more complexity to the build. > > > > Another option that I am considering is to discontinue support for JDK 8 > > and only support JDK 11+ to be able to remove quite some build-logic. > > > > > > Regards... Dominik. > > > > > > On Tue, May 9, 2023 at 5:59 PM PJ Fanning <fannin...@yahoo.com.invalid> > > wrote: > > > > > Hi everyone, > > > > > > The CI builds are passing with newer Java versions (11 and 17) but > failing > > > with Java 8. The builds seem to crash and timeout. > > > > > > https://ci-builds.apache.org/job/POI/ > > > > > > I can't reproduce the issues on my laptop. > > > > > > I raised https://issues.apache.org/jira/browse/INFRA-24362 but didn't > get > > > much info from it. > > > > > > Anyone have any ideas on how to proceed? > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: dev-unsubscr...@poi.apache.org > > > For additional commands, e-mail: dev-h...@poi.apache.org > > > > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@poi.apache.org > For additional commands, e-mail: dev-h...@poi.apache.org > > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@poi.apache.org For additional commands, e-mail: dev-h...@poi.apache.org
