RFR: JDK-8261034: improve jcmd GC.class_histogram to support parallel

Lin Zang lzang at openjdk.java.net
Fri Feb 12 01:07:38 UTC 2021


On Thu, 11 Feb 2021 22:50:50 GMT, Chris Plummer <cjplummer at openjdk.org> wrote:

>> Dear All, 
>> I took the liberity to create a PR (#2519) for the "noparallel" option. Maybe we could discuss there.  And I also found @Hamlin-Li has created a CSR for that. Thanks a lot!
>> 
>> BRs,
>> Lin
>
> There is still one minor issue with this `noparallel` support, and that is we still run into the problem of `jmap -dump` not being able to pass a 4th argument via the attach API. We could just say that's fine, and require using the `jcmd` to do the heap dump if the user wants to override the default of having parallel enabled, but then it seems we should do the same for `jmap -histo` to be consistent. So in other words, no support in `jmap` for disabling parallel support, but we do have the support in `jcmd`. Thoughts?

Hi Chris, 

> There is still one minor issue with this `noparallel` support, and that is we still run into the problem of `jmap -dump` not being able to pass a 4th argument via the attach API. We could just say that's fine, and require using the `jcmd` to do the heap dump if the user wants to override the default of having parallel enabled, but then it seems we should do the same for `jmap -histo` to be consistent. So in other words, no support in `jmap` for disabling parallel support, but we do have the support in `jcmd`. Thoughts?

I just updated the #2261 which introduce a new command "dumpheapext" that could handle more arguments (as you suggested the idea :-> ). I think maybe that kind of change is acceptable. Moreover, this way also allow us to add more options in future.

So maybe we have two choice here:
- Add a new command for argument extension.   so jcmd and jmap could be consisitent.
- Remove the "parallel" option in jmap, and leave the control to jcmd.

I prefer the first one as it may be more clearer for users to have consisitent usage of jcmd and jmap tools as they were before. 

Thanks!
Lin

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

PR: https://git.openjdk.java.net/jdk/pull/2379


More information about the serviceability-dev mailing list