Fabian Ruffy

@jafingerhut Do you know whether there are any specific restrictions on default action instantiations on tables? Otherwise, we could just create a simple P4Info message that describes an action call....

I was thinking many of the features of a write request are not required here. ``` message TableAction { oneof type { Action action = 1; uint32 action_profile_member_id = 2;...

Sounds good to me! I can give this a crack. I can also reuse your descriptions for the documentation.

> @fruffy bump - can we do this by end of month to meet 1.4.0 release goals? Thx. Currently have a lot on my plate. Will do my best but...

> @fruffy Thanks for expediting this! It looks good to me, but before I "approve" I have a question. We often like to have a p4c front-end validation (or at...

> LGTM, thanks! How's your merge process for these things? Should I just go ahead and merge or wait for @smolkaj?

Could this help resolve some diamond dependency problems we have? (e.g., p4runtime vs p4-constraints vs p4c dependency conflicts) Or fix https://github.com/p4lang/p4c/pull/4640 where we can not export some feature flags?

The issue is effectively: https://github.com/abseil/abseil-cpp/issues/740 We are working around it by setting some [options](https://github.com/p4lang/p4c/blob/main/cmake/Abseil.cmake#L62) in CMake, but it would be nice to be consistent. Abseil is the only third-party system...

If I'd wager a guess it might be because of the Tofino IR sources: https://github.com/p4lang/p4c/blob/main/backends/tofino/CMakeLists.txt#L219 which need to be compile with the ir-generated blob.

>But looks like these are not entirely frontend... Yeah I also think they could be cleaned up. There includes are quite messy and there are some circular dependencies.