C++: Fix IR variable reuse for global var inits

[rdmarsh2]

Fixes a performance issue where all IRVariables corresponding to a given global variable were being used in the initializer IRFunction for that global variable.

[github-advanced-security]

Found 1 vulnerability.

[github-advanced-security]

Found 1 vulnerability.

[dbartol]

LGTM as long as the tests pass.

[jketema]

I fixed the test regressions caused by running against the old frontend. However, it turns out, also some syntax zoo tests broke (unrelated to the frontend). I could update those, but I cannot tell if the changes are correct.

Also the IR VariableAddresses without names look funny especially for declarations like:

int global_6 = global_2;

This becomes:

# 1769| int global_6
# 1769|   Block 0
# 1769|     v1769_1(void)       = EnterFunction     : 
# 1769|     mu1769_2(unknown)   = AliasedDefinition : 
# 1769|     r1769_3(glval<int>) = VariableAddress   : 
# 1769|     r1769_4(glval<int>) = VariableAddress   : 
# 1769|     r1769_5(int)        = Load[?]           : &:r1769_4, ~m?
# 1769|     mu1769_6(int)       = Store[?]          : &:r1769_3, r1769_5
# 1769|     v1769_7(void)       = ReturnVoid        : 
# 1769|     v1769_8(void)       = AliasedUse        : ~m?
# 1769|     v1769_9(void)       = ExitFunction      : 

Which VariableAddress is which here?

[rdmarsh2]

Also the IR VariableAddresses without names look funny especially for declarations like:

Fixed by d1c6022

[jketema]

Fixed by d1c6022

Thanks. Would it be possible to add int global_6 = global_2; as a test?

[rdmarsh2]

Good call - I've added the test and fixed another case of missing variables

[github-advanced-security]

Found 9 vulnerabilities.

[github-advanced-security]

Found 1 vulnerability.

[MathiasVP]

Everything LGTM! Do we need to rerun DCA? I don't think much has changed apart from a merge-from-main and accepting some test changes, right?

[rdmarsh2]

I don't think we do... My only worry would be how much change there's been to main in the meantime, but there hasn't been much else happening on the C++ dataflow side