RFR: JDK-8318636: Add jcmd to print annotated process memory map [v9]

Thomas Stuefe stuefe at openjdk.org
Fri Nov 10 07:41:05 UTC 2023


On Thu, 9 Nov 2023 07:03:34 GMT, David Holmes <dholmes at openjdk.org> wrote:

>>> > As this adds a JCmd, doesn't this need both a CSR and a manual entry?
>>> 
>>>     * CSR: not sure; there are precedences for going with CSR and without CSR. Since we get awfully close to JDK22 freeze, I would prefer for a CSR not to be needed. Also would make backporting a lot easier.
>>> 
>>>     * Manpage: not sure either; IIRC last time I tried, I was told that Oracle maintains these internally, because there is internal documentation that needs updating too?
>> 
>> @dholmes-ora, would you mind helping us out here with regards to the process? Would a new JCmd require a CSR or would it be acceptable to merge without one? Thank you.
>
>> > > As this adds a JCmd, doesn't this need both a CSR and a manual entry?
>> > 
>> > 
>> > ```
>> > * CSR: not sure; there are precedences for going with CSR and without CSR. Since we get awfully close to JDK22 freeze, I would prefer for a CSR not to be needed. Also would make backporting a lot easier.
>> > 
>> > * Manpage: not sure either; IIRC last time I tried, I was told that Oracle maintains these internally, because there is internal documentation that needs updating too?
>> > ```
>> 
>> @dholmes-ora, would you mind helping us out here with regards to the process? Would a new JCmd require a CSR or would it be acceptable to merge without one? Thank you.
> 
> @jdksjolen No CSR needed: from another related PR - "We do not use CSR requests with jcmd changes as they are deemed diagnostic commands - ref JDK-8203682"
> 
> But yes the `jcmd` manpage should be updated ref:
> https://docs.oracle.com/en/java/javase/19/docs/specs/man/jcmd.html
> though I'm worried we may not have kept it up to date! That requires an Oracle engineer to apply the changes to the jcmd.md markdown sources in our repo, and then regenerate the `jcmd.1` manpage file. The doc update can be split into a separate doc sub-task so that the main PR is not held up. (And we probably need an audit to see if any other updates are missing - which is painful.)

@dholmes-ora thank you for your review. I hope I have addressed all concerns. I made this all Linux only, removed os::realpath, print errno.

I am currently working on a follow-up RFE that provides this command for Windows, and it works, but I will do it in a separate RFE (rolling back the ifdef LINUX should be easy) and possibly not for JDK 22. https://github.com/openjdk/jdk/pull/16593

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

PR Comment: https://git.openjdk.org/jdk/pull/16301#issuecomment-1805238529


More information about the serviceability-dev mailing list