Allow reuse of typenames as normal names in the parser, as long as it…

[ChrisDodd]

… does not lead to ambiguities

This revealed an error check that wasn't being done (relying on a syntax error to catch the error).

I'm not certain, but I think that some updates are needed to the ProgramStructure object to properly track names that are redefined as non-types, as it tries to maintain a separate data structure rather than using the program's symbol table.

[mihaibudiu]

ProgramStructure is only used for P4-14 programs.

[ChrisDodd]

Naming confusion here -- there's a P4V1::ProgramStructure (used for P4-14
programs) and an (unrelated) Util::ProgramStructure, used to track whether
identifiers should be TYPE, NAMESPACE or simple IDENTIFIER tokens. I was
referring to the latter.

On Mon, Oct 3, 2016 at 1:00 PM, mbudiu-vmw notifications@github.com wrote:

ProgramStructure is only used for P4-14 programs.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#97 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/AD4c8UYrUjAslFfY3w2MJNvojcnEze9Lks5qwV7WgaJpZM4KM9Go
.

[mihaibudiu]

I will need to write a few test cases to figure out whether this works.

[mihaibudiu]

By reading the code in p4-parse.ypp, it looks like named namespaces are never used.
The code was designed to accommodate a namespace X { } construct, which is not implemented (or part of the spec).
The p4-parse file creates namespaces in two places only, and they all have an empty string name, only while parsing method arguments.
So I don't think that ProgramStructure can ever return NAMESPACE as the syntax is today. It certainly does not for any of our test cases.
So I believe this patch can be merged.
There's a secondary question whether we want to simplify the grammar and ProgramStructure.