Relocate the J2I virtual thunk in PIC data on x86

[jdmpapin]

No description provided.

[jdmpapin]

Marked WIP because this depends on both #2851 and #2874. For the moment I have copies of those changes in this branch so that it will build, but this PR is just for the change in d281f6b

[jdmpapin]

@irinarada FYI

[mstoodle]

The may be reasonable reasons why what I've suggested is hard to do, and if we're just going to change the structure of these relocations to eliminate this "offset" value anyway, then it may not be worth fixing. But this is one of those things about AOT relocations that generally comes back to bite painfully (I speak from experience, sadly).

[irinarada]

@jdmpapin - do we have an ETA for this one?

[jdmpapin]

Rebased to pick up #2649 for Travis, and removed WIP now that the dependencies have been merged

[andrewcraik]

Jenkins test xlinux,win jdk11,jdk8

[andrewcraik]

Jenkins test sanity xlinux,win jdk11,jdk8

[andrewcraik]

Jenkins test sanity xlinux,win jdk11,jdk8

[jdmpapin]

The line endings check timed out, likely not due to line ending problems. Can we restart it?

Linux JDK11 sanity had a few failures, none of which are related to my changes:

• test_toLowerCase_lithuanian is Test-sanity.functional-JDK11-linux_390-64_cmprssptrs JCL_Test_0 test_toLowerCase_lithuanian #2890;
• CloneWeakReferenceTest is CloneWeakReferenceTest failed with CloneNotSupportedException on linux_x86-64_cmprssptrs in JDK11 #2801; and
• cmdLineTester_SCCommandLineOptionTests_1 nameOption6 and nameOption18 were due to 528b82f, which has been reverted.

Windows JDK11 sanity failed test_toLowerCase_lithuanian and CloneWeakReferenceTest, as above.

Linux JDK8 sanity failed nameOption6 and nameOption18, as above.

[fjeremic]

#2890 is fixed. #2801 is in progress. I'll let whoever gets assigned to this evaluate whether we should merge or wait for a clean build. Probably @andrewcraik?

Jenkins line endings check

[andrewcraik]

given that this is core feature work for 0.10, that the tests in question will not be affected by the changes that are intended to be made in this change, and that we need to complete feature work for 0.10 I'm going to say that I am happy merging this at this time.

Thanks all for the detailed analysis of the results.

[andrewcraik]

LGTM