Open XL header includes and vm flag fixup

[Deigue]

runtime/tests/jsig/main.c : dlfcn.h needs to be defined before defines for dllfree()/dllload() due to conflicting type definitions decomp003.c/decomp005.c : strcasecmp() implict function declaration needs strings.h j9vmemTest.c : -Wimplicit-int , missing type specifier that defaults to "int"
jvminit.c : atoe_enableFileTagging() implicit function declaration requires atoe.h
runtime/vm/CMakeLists.txt : Added appropriate conditionals to apply correct flags for Open XL suggested by compiler team. Hooks are not implemented for Open XL.

[r30shah]

I just have one question I have asked, changes LGTM. Will launch test internally with this to make sure we do not run into any issue before continuing to merge.

[r30shah]

runtime/vm/CMakeLists.txt : Added appropriate conditionals to apply correct flags for Open XL suggested by compiler team. Hooks are not implemented for Open XL.

Just to understand, with xlc it removes -O3, -g for compiler but with OpenXL it will not ?

[joransiu]

Why won't we want -O2 being added for interp_new_flags from the XLC case?

[Deigue]

That might be an oversight, from the discussions I can find right now the intention was only to remove the "\"-Wc flags , but leave the same optimizations. So I will try with restoring the optimization levels and rebuilding.
If there is something I may not be remembering, at least it might remind me if it was intentional.

[Deigue]

I think this is what the intended outcome was, which I am currently testing out ...

    if(OMR_TOOLCONFIG STREQUAL "openxl")                                                                                                                                                                                   # Hooks are not implemented in Open XL. The default debug compiles in Open XL will be NOHOOK
        # equivalent.
        list(APPEND interp_flags_to_remove "-O3" "-g")
        list(APPEND interp_new_flags -O2)
    else()
        # TODO need to handle optimized debug info
        list(APPEND interp_flags_to_remove "-O3" "-g" "-qdebug=nohook")

        list(APPEND interp_new_flags
            -O2
            "\"-Wc,TBYDBG(-qdebug=MRABIG)\""
            "\"-Wc,TBYDBG(-qdebug=lincomm:ptranl:tfbagg)\""
            "\"-Wc,FEDBG(-qxflag=InlineDespiteVolatileInArgs)\""
        )    

[Deigue]

Verified these changes are working with openxl, still need to test xlc with jenkins, but I imagine it won't be a problem since that clause/conditional has remained unchanged altogether.

[r30shah]

@Deigue We may need this guarded by platform - See failures on other platform (/job_output.php?id=89250736)

[Deigue]

Done

[Deigue]

I have removed runtime/tests/jsig/main.c reordering of headers as per @joransiu's advice. I discussed with Keith and have integrated this dependency where it should be (in the a2e wrappers, see dll.h change in eclipse-omr/omr#7383 , explicitly defined dlfcn.h as a pre-req for the functions it may override with its wrapper.

edit: Also rebased branch to eclipse/master

[Deigue]

jdk21 build passes , same commit based off ibm-openxl-headers (internal fork)

[Deigue]

This one is ready @r30shah , and the dependent changes from eclipse-omr/omr#7383 fixing the headers has been merged.

EDIT: Sorry, will update this once I have verified Java 8 testing, need to verify that as well.

[Deigue]

All passed with java 8 and java 21 builds.

[r30shah]

jenkins test sanity zlinux jdk21

[r30shah]

Failing test in the launched test is same as #19618, not related to change in this PR. Merging this as internal tests have been confirmed by @Deigue and changes have been approved.

[r30shah]

@Deigue Given that this is causing failure - I wonder, we should guard this with J9ZOS same as how atoe.h was done (Assuming that only z/OS needs the function defined in strings.hwhich you may have bumped into with OpenXL work.

[Deigue]

yes that sounds like a good idea, I think the build must have failed on one of the platforms we didnt verify with.
Same for decomp005.c too I guess...