Add selected tracepoints for JCL

[pshipton]

Issue https://github.ibm.com/runtimes/semeru-requests/issues/46

[keithc-ca]

This deserves a comment explaining why a header file includes a C source file.

[keithc-ca]

On second thought, if including tracehelp.c works, perhaps it should just appear in the .c file that includes jcltracepoints.h? It still feels wrong to include .c files in a .h file.

[pshipton]

If I move tracehelp.c out, then jcltracepoints.h is no longer specific to tracepoints, but more generic for loading j9.h outside of OpenJ9. jcltracepoints may no longer be an appropriate name, does includej9h.h work for you?

[keithc-ca]

That seems somewhat awkward to me, but I'm not sure I have a better suggestion.

More troubling is the collection of configuration definitions: I think we should be including j9cfg.h instead (a number of source files in the extensions repos already do that).

[pshipton]

Are you talking about the parts like the following? I don't believe including j9cfg.h fixes these. The OpenJ9 definitions are on the command line, not defined in j9cfg.h afaik. I need to translate OpenJDK platform definitions to OpenJ9 definitions. Even j9cfg.h won't define the right things unless these are set.

#if defined(AIX)
#define AIXPPC
#define RS6000
#endif /* defined(AIX) */

[pshipton]

Also note that j9.h does include j9cfg.h

[keithc-ca]

I'm not sure includej9.h is the best name for this header file; perhaps j9internals.h or j9access.h (hinting that it provides access to internal details of J9). Whatever we decide, the name of the macro here should be derived from the file name.

[pshipton]

Renamed to j9access.h

[keithc-ca]

It's not clear to me that any of this should be necessary.

[pshipton]

They were all added to fix compilation errors. Do you need me to retry builds without them and record the errors that occur? OpenJ9/OMR use different platform defines than OpenJDK uses/defines. If no matching OpenJ9 platform define is set, there are a couple places that define structures based on platform, and then a compilation error occurs because the structure isn't defined. That was the problem for AIX/Mac, for Windows it may be slightly different but along the same lines.

[pshipton]

The errors when OSX is not defined below. I think they were the same for AIX but both RS600 and AIXPPC are needed to fix both of these.

/Users/jenkins/peter/openj9-openjdk-jdk8/openj9/runtime/oti/j9port.h:82:2: error: "J9PORT_LIBRARY_SUFFIX must be defined"
#error "J9PORT_LIBRARY_SUFFIX must be defined"
 ^
/Users/jenkins/peter/openj9-openjdk-jdk8/openj9/runtime/oti/j9port.h:324:2: error: unknown type name 'SYS_FLOAT'
        SYS_FLOAT cpuEntitlement;

[pshipton]

On Windows without OMR_OS_WINDOWS:
c:\workspace\openj9-openjdk-jdk8\omr\include_core\omrport.h(55): fatal error C1083: Cannot open include file: 'unistd.h': No such file or directory