Fixing bug for ParserUnroll application to a p4 program with a header union #3121
Conversation
midend/parserUnroll.cpp
Outdated
| if (element->is<IR::ArrayIndex>()) { | ||
| const IR::Expression* left = element->to<IR::ArrayIndex>()->left; | ||
| if (left->type->is<IR::Type_Stack>()) | ||
| return left->type->to<IR::Type_Stack>()->elementType; | ||
| } | ||
| return typeMap->getType(element, true); | ||
| auto* currentType = typeMap->getType(element); |
There was a problem hiding this comment.
How can it happen that type information is missing here? Is this a problem with type inference?
There was a problem hiding this comment.
The problem with right part of the assignment. If element is hdr.h[0].i1.index then typeMap->getType(element) returns nullptr.
|
The right solution is to fix the evaluation of expressions with type headerunion in the symbolic interpreter. |
The getTypeArray function is not a part of the interpreter and It doesn't take part in a calculation process. #3063 contains changes for support header union in a evaluation process. I think that is a some particular case for support of header union. |
|
Now that #3063 has been merged what is the status of this pull request? |
|
@mbudiu-vmw Do you have any comments on this pull request? |
|
Yes, I have requested changes which have not been carried out. |
Could, you, please, specify them? |
|
What is the status of this PR? |
|
As far as I understood we are waiting for solving the issue with the description of header union in p4lang here. It was opened two years ago. It looks like we are stuck here. |
|
The compiler should usually be ahead of the spec - this helps us make sure that the spec is right. |
|
@mbudiu-vmw, Could you, please, make a review again? |
|
This PR contains only test cases, is this intended? |
I would like to add an additional test for header union stacks. |
These small changes allow to use ParserUnroll class for the programs with header unions.