Calling ProcessLine in an unnamed macro leave Cling in a broken state .. sometimes about root

[hahnjo]

I personally don't think this should block the 6.32 release: As far as I understand, this has basically always been broken in Cling, and I verified that constRetrunTest_WILL_FAIL doesn't pass either on 6.20 using LLVM 5:

1079: Processing /home/jhahnfel/ROOT/v6-20-00-patches/src/roottest/cling/return/run.C...
1079: /home/jhahnfel/ROOT/v6-20-00-patches/src/roottest/cling/return/run.C:7:4: error: expected ';' after expression
1079: t02 t;
1079:    ^
1079:    ;
1079: /home/jhahnfel/ROOT/v6-20-00-patches/src/roottest/cling/return/run.C:7:5: error: use of undeclared identifier 't'
1079: t02 t;
1079:     ^
1079: /home/jhahnfel/ROOT/v6-20-00-patches/src/roottest/cling/return/run.C:8:15: error: use of undeclared identifier 't'
1079: TObject o = **t.getVal();
1079:               ^
1079: /home/jhahnfel/ROOT/v6-20-00-patches/src/roottest/cling/return/run.C:9:24: error: use of undeclared identifier 't'
1079: TObject * const * op = t.getVal();
1079:                        ^

[pcanal]

Will_Fail is the (now) expected result and lead to a 'succesfull' run of the test. What is "new" is that there is one platform where this failure is now a segmentation fault of sort.