RFR: 8193513: add support for printing a stack trace on class loading [v7]

[Ioi Lam]

On Fri, 7 Jul 2023 15:27:04 GMT, Doug Simon <dnsimon at openjdk.org> wrote:

>> src/hotspot/share/oops/instanceKlass.cpp line 3749:
>> 
>>> 3747: 
>>> 3748:   print_class_load_cause_logging();
>>> 3749: 
>> 
>> This prints the stacks before printing the name of the class. I think the name should be printed first.
>
> Not sure what you mean. The `class+load+cause` log output includes the name before the stack. See the updated examples in this PR's description.
> 
> If you mean `class+load+cause` log output precedes `class+load` log output, then yes, that is the case and it's intentional to keep these logging features independent of each other as David suggested (https://github.com/openjdk/jdk/pull/14553#discussion_r1249847278). Since they are independent, it should not matter which output comes first.

Even if they are independent, relevant output should be grouped together. Currently, if we specify `-Xlog:class+load*=debug`, the output would look like this:


Java stack when loading  java.util.concurrent.ConcurrentHashMap$ValuesView
...
[many many lines of output]
...
java.util.concurrent.ConcurrentHashMap$ValuesView source: shared objects file
  klass: 0x000000080033a0f8 super: 0x000000080033a320 loader: [loader data: 0x00007faa34113890 of 'bootstrap']

This makes it difficult to see all the information of this class at a glance.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/14553#discussion_r1256041975