Allow annotations on functions

[ChrisDodd]

Probably an oversight that this wasn't allowed before. Use case is to specify which hardware resource(s) should be used to implement a specific function

[jafingerhut]

Good addition. Can you add a simple test program using this feature, too?

[jafingerhut]

Do you want to create an issue and/or PR for the language spec, too? If you would prefer somone else do that, I can do it.

[ChrisDodd]

Do you want to create an issue and/or PR for the language spec, too? If you would prefer somone else do that, I can do it.

The spec is a bit unclear, but seems to allow annotations on any 'element' in a program, but it is not clear what is and is not an element. Besides this case, p4c currently does not allow annotations on error and match_kind, nor on any statement except for block statements (allowing them on statements that start with a ( would be problematic, parsing wise).

[jafingerhut]

Do you want to create an issue and/or PR for the language spec, too? If you would prefer somone else do that, I can do it.

The spec is a bit unclear, but seems to allow annotations on any 'element' in a program, but it is not clear what is and is not an element. Besides this case, p4c currently does not allow annotations on error and match_kind, nor on any statement except for block statements (allowing them on statements that start with a ( would be problematic, parsing wise).

Agreed that the spec English prose can be a bit unclear on this point. The part that would be worth updating in the spec is the grammar.mdk file that is currently almost the same as the Bison grammar used by p4c. This PR updates the grammar used by p4c, so making corresponding changes in the grammar.mdk file of the spec would be good.

[ChrisDodd]

Good addition. Can you add a simple test program using this feature, too?

I can add a testcase defining a trivial function with a random annotation, but that really only tests that the perser accepts it (which is pretty trivial -- not sure if it is worth the extra 1 second of CI time)

[ChrisDodd]

I added a spec (grammar) change to match #1274

[jafingerhut]

LGTM, although I am not the best person to review the C++ parts of the changes, which is most of this :-)