RFR (M) 8223456: CSR Delayed starting of debugging via jcmd

[Langer, Christoph]

Hi Alan,

> On 27/05/2019 14:23, Schmelter, Ralf wrote:
> 
> I need reviews for the following CSR:
> https://bugs.openjdk.java.net/browse/JDK-8223456
> Note that this CSR is for a feature which already was commited to JDK 12.
> 
> I think this feature needs to be re-examined, maybe backed out or the
> onjcmd sub-option hidden from the help output if we can't get agreement
> before JDK 13 RDP1. It's unfortunate that it was pushed to JDK 12 without a
> CSR because that would have been the opportunity to ask questions. The
> problem with this feature is that it ties the debugger agent to a unrelated
> JDK-specific tool. It feels like it conflates two things.  So I think the "feature"
> needs to be re-discussed to see whether the scenario is really compelling
> and to see the list of alternatives that were explored.

I don't fully understand what you are saying, especially the part "it ties the debugger agent to a unrelated JDK-specific tool".

The principal feature here is that one wants to be able to "debug on demand". Currently (or before the change that we're discussing was implemented) one had to start the JVM with the JDWP agent activated in case one's planning to debug. The "onjcmd" option gives the possibility to start the JDWP agent in a standby mode and later on activate the actual debugging. The most common way will probably be to use jcmd but there are also other ways using MXBeans to trigger the debugging. So I don't see the tie to a certain JDK-specific tool here.

I think the overall story of "debugging on demand" is quite compelling. We've had that for years in SAP's proprietary JVM and it was well received by the users. It gives you the option to connect with the debugger post mortem to analyze issues.

Anyway, I have reviewed Ralf's CSR and we've put it to status "Proposed". We're open for discussion, of course, on how to optimally implement this in OpenJDK. E.g. I personally don't think the naming of the option "onjcmd" is a good choice. And there's probably more around this feature which we'd need such as means to get the listen address of the debugger (or to get the JDWP agent's configuration, reconfigure it or disarm it). Maybe this can all be part of the CSR (or an upcoming CSR). Maybe it's even a candidate for a JEP, I don't know. So let's use the CSR to discuss this.

By the way, I had asked at the time the change was reviewed, whether we needed a CSR [0] but didn't get a response. Sometimes it's not easy to get people involved and a fruitful discussion started, like the one you are obviously wishing for...

Best regards
Christoph

[0] https://mail.openjdk.java.net/pipermail/serviceability-dev/2018-December/026360.html