Prevent javacore generation from interfering with stack walker

[gacholio]

Add a flag to getOwnedObjectMonitors to prevent the stack walker from asserting and marking the thread stack corrupt.

Also remove the private flag for stack corruption - it has served no purpose for many releases.

Internal 150400, 150750.

[gacholio]

@keithc-ca Please verify that my removal of the private flag won't break DDR.

[gacholio]

This is related to RTC story 150400.

[keithc-ca]

Please verify that my removal of the private flag won't break DDR.

I don't see any potential problems for DDR.

Please identify the assertion(s) that will be avoided by this change.

[gacholio]

Please identify the assertion(s) that will be avoided by this change.

It avoids the "invalid JIT return address" assertion if the stack walker is called from getOwnedObjectMonitors with the reportErrors flag FALSE (at the moment, only javacore generation calls this way).

[keithc-ca]

I don't see any potential problems for DDR.

Actually, I take that back. We should keep J9_PRIVATE_FLAGS_STACK_CORRUPT (perhaps only as a (reverse) compatibility constant in CompatibilityConstants29.dat) so it is properly recognized in old core files.

Add to CompatibilityConstants29.dat:

J9Consts.J9_PRIVATE_FLAGS_STACK_CORRUPT = 0 /* no longer used */

and don't change DDR_VM/src/com/ibm/j9ddr/vm29/j9/stackwalker/StackWalker.java.

[gacholio]

The tag bit is meaningless - stacks are never corrupt when running, and the tagging of foreign threads in the javacore writer is both unsafe an incorrect. DDR should not be processing this flag at all.

[keithc-ca]

If the flag is set - even if it was in error - DDR should not ignore it.

[gacholio]

No - the only possible result of the check is that the stack walk will fail in DDR (assuming any code in DDR even checks the return value), which serves no useful purpose. Hiding a possibly bad stack from the debug tools is the precise opposite of what we want.

[keithc-ca]

Hiding a possibly bad stack from the debug tools is the precise opposite of what we want.

Agreed. One can use the !threads command to see privateFlags if necessary.

[keithc-ca]

Jenkins test sanity alinux jdk17