On Fri, 28 Aug 2026 08:28:20 GMT, Alan Bateman <[email protected]> wrote:
>> Coleen Phillimore has updated the pull request incrementally with one >> additional commit since the last revision: >> >> Fix copyrights. > > @coleenp Would it be possible to provide a summary on the how this changes > HRPOF heap dumps and JVMTI heap walking tools? > > I'm wondering about tools that use functions like FollowReferences to follow > refs. If a ClassLoader object is encountered today then it will follow the > refs to all classes defined by that class loader. With the change then it > won't do that, is that right? I'm wondering if the refs will need to be > synthesized to avoid breakage. > @AlanBateman you are right that heapdump and followrefs tools will not visit > the class after visiting the class loader with this change. They'll still > visit the class loader with the class. I'm still trying to work out how to > describe this change in behavior for these tools and maybe have a CSR for it. > There was only one test left that failed because of this change - was hoping > some hprof test would fail so I could see the effect of that. I'm still > working on this though, but I wanted to publicize the change first. I worry this is a significant behavior change that will blind some heap analysis tools. As part of "I'm still working on this", would you mind looking synthesizing these refs so that tools using FollowReferences will follow refs to the loaded classes? ------------- PR Comment: https://git.openjdk.org/jdk/pull/32519#issuecomment-5452693783
