RFR: 8328866: Add raw monitor rank support to the debug agent. [v6]

[Chris Plummer]

On Mon, 13 May 2024 21:47:32 GMT, Alex Menkov <amenkov at openjdk.org> wrote:

> If debugger attaches to the debuggee, detaches and re-attaches again, are the monitors recreated again?
> (with rankedMonitor you added an assert if `DebugRawMonitor::monitor` is not null)

Almost all of the debugMonitorCreate() calls seem to come (indirectly) from initialize(), which is only called once.

cmdQueueLock, which is the one monitor we call debugMonitorDestroy() on, is called from debugLoop_run(), which is called whenever a debugger connection is initiated. This is the only case of creating, destroying, and then recreating. I wonder if there is some need for that.

popFrameEventLock and popFrameProceedLock are created lazily when first needed. There is a check to make sure they are not recreated if already created. They are never destroyed.

That seems to be it. So we pretty much create all monitors upon initialization and never recreate those that are created during initialization.

With regard to the assert, this can only be an issue for cmdQueueLock, and since  debugMonitorDestroy() is called on it (and that clears the needed fields), the assert shouldn't be triggered.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/19044#discussion_r1599160098