Segmentation fault error during federated execution caused by multiport

[jackyk02]

This pull request addresses the issue found in the convert_C_port_to_py function. The problem occurs when handling generic_port_instance_struct* cport for multiport instances. Currently, this structure is not properly assigned when dealing with multiport instances, which leads to subsequent issues in the function. On line 433, we see the macro FEDERATED_ASSIGN_FIELDS(((generic_port_capsule_struct*)cap), cport), which relies on cport. In the case of multiport, if cport isn't appropriately assigned, this macro won't function correctly, causing segmentation fault.

[petervdonovan]

I left a little question/quibble, but basically I am just happy that it works now. Thank you Jacky! this might have saved me a ton of time.

[petervdonovan]

I wonder if the intended design was to keep this as-is, and to wrap line 433 (FEDERATED_ASSIGN_FIELDS) in if (width == -2). Just guessing since it seems like cport is only being used when the port is not a multiport.

In fact it might have been clearer to move all of the code that touches cport into the branch at line 435-448.

[edwardalee]

I agree with Peter here. The proposed cast doesn't look correct for a multiport, and, although it is probably harmless, it seems misleading.

[edwardalee]

Good diagnosis! I agree with @petervdonovan that there may be a slightly better fix.

[edwardalee]

I agree with Peter here. The proposed cast doesn't look correct for a multiport, and, although it is probably harmless, it seems misleading.

[lhstrh]

@petervdonovan, if you could implement your suggested change and file a PR in reactor-c, that would be great!

[petervdonovan]

Closing as replaced by lf-lang/reactor-c#218