On Wed, 26 Aug 2026 19:19:13 GMT, Chris Plummer <[email protected]> wrote:

> This fixes the issue with the heap walk bailing when encountering a 
> StackChunk, but what actually gets dumped into the hprof file when a 
> StackChunk is encountered?

Based on my understanding the chunk itself is written as a regular instance 
record with just its declared fields, parent, sp, size and so on. the copied 
stack payload isn’t a java field so it doesn’t appear in the record, the size 
override only makes the walk step over it correctly so everything after the 
chunk gets dumped. the frames’ references still aren’t emitted as outgoing refs 
from the chunk, that’s the scanning change coming separately on the old sa 
stack chunk bug. so the dump now contains every heap object, but something held 
only by a mounted vthread stack shows up without that incoming reference until 
the scanning lands

-------------

PR Comment: https://git.openjdk.org/jdk/pull/32525#issuecomment-5430266723

Reply via email to