Store classes not worth remembering per client for JITServer

[dmitry-ten]

PR #10644 added a static array of classes not worth remembering
to Symbol Validation Manager. JITServer, however, requires this array
to be stored on a per client bases, since pointers to J9 classes
are different between clients.

[dmitry-ten]

@mpirvu this should fix #10663. I started the tests, will update when they complete.

[mpirvu]

It looks mostly ok. Please see the inline suggestions

[mpirvu]

Not sure if this can be used to determine the number of elements in the array, since the array is defined in cpp while we need this definition in hpp:

template <typename T>
constexpr auto sizeof_array(const T& iarray) 
{
    return (sizeof(iarray) / sizeof(iarray[0]));
}

[dmitry-ten]

@mpirvu other than replacing vector with array, I addressed all comments.
Replacing vector with array requires creating a dynamically allocated array though, which needs default constructor. And to assign to that array, assignment operator is also needed. I'm not sure if the additional code would be worth it.
@dsouzai could you also take a look?

[mpirvu]

LGTM

[mpirvu]

We could use an embedded array if we knew the size. If the size of the original array cannot be statically determined, an workaround is to declare it as '2' in the hpp file and then to add a check at compile time that the number of entries in the cpp is actually 2. I'll let @dsouzai decide if he is ok with that change

[dsouzai]

Yeah looks ok to me.

If the size of the original array cannot be statically determined, an workaround is to declare it as '2' in the hpp file and then to add a check at compile time that the number of entries in the cpp is actually 2.

Yeah I'm ok with this since it'll be a compile time static assert, so it will at least get caught pretty early on.

[dmitry-ten]

@mpirvu changed vector to static array and removed the fix to front-end for getSystemClassFromSignature message.

[mpirvu]

Can't we leave the first const here. The second one refers to the _className being immutable, but the first one refers to the string of characters being immutable.

[dmitry-ten]

True. I can push the change once the tests pass.

[mpirvu]

The tests on Power take very long for some reason, but on x86 they passed. You could do this change now.

[mpirvu]

jenkins test sanity plinuxjit,xlinuxjit jdk11

[dmitry-ten]

Pushed

[mpirvu]

jenkins test sanity plinuxjit,xlinuxjit jdk11

[mpirvu]

Testing on x86 was successful. Testing on Power seems to hand for some reason even though all the tests have passed. I don't know what the infra is waiting for.
I am going to merge this item since it addresses some serious breakage for JITServer.