RFR: 7033: Allow overriding the id (name) of the event type

[Gunnar Morling]

On Wed, 2 Dec 2020 21:32:08 GMT, Marcus Hirt <hirt at openjdk.org> wrote:

>>> and it will simply use the class name as event name if not specified (IIRC)
>> 
>> Yes, that's what was (is) doing before this PR. I don't think it's a good idea though, as the event type is located in the package of the instrumented code, and thus this event name also would be related to that, which I think is confusing.
>> 
>> So I changed it to base the event name of the name given in the XML file (which you just renamed to "label", which makes sense to me). So I could see us doing either what I propose here, or derive the name from the id given in the XML (makes more sense to me now in the light of that renaming you did). And perhaps even make id mandatory in the XSD?
>
> Yes - the name should be derived from the id, if set (that's why I fixed it up in those other PRs - so that you could use it here properly). :)
> 
> IMHO, ID should not be mandatory in the XSD - the default should be the default in JFR, which is to just use the class name - which admittedly will be a bit bonkers if generated by the agent. But if you don't care at all about the class name, because you're just doing a one off, throwaway, kinda thing, getting the generated one is fine, at least to me. The labels is all you care about then anyways. Now, if you are to be a good citizen, you'll provide an ID, and all our examples should have them. 
> 
> We may also want to validate that all registered IDs are unique, and warn and exit if we find duplicates.

Ok, cool. I'll update this PR along those lines then.

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

PR: https://git.openjdk.java.net/jmc/pull/170