From: Chris Torek <chris.torek@gmail.com>
Date: Wed, 11 Nov 2020 20:34:59 GMT
Subject: Re: [PATCH v5 12/15] strmap: enable allocations to come from a mem_pool
Message-ID: <CAPx1GvfXMnhgQDan7N=Y4WWdcS6p6WrekYkDyfx4nwnDtSuMQQ@mail.gmail.com>
In-Reply-To: <20201111190121.GE9902@coredump.intra.peff.net>

On Wed, Nov 11, 2020 at 11:02 AM Jeff King <peff@peff.net> wrote:
> Even if we're only using a pointer to it, we still need a valid forward
> declaration [in a function, but not in a struct definition] ...
> I'm not sure whether this is a seldom-seen corner of the C standard, or
> a compiler-specific thing (though both clang and gcc seem to allow it).

It is standard (and not all *that* seldom-seen, in that many compilers
have warnings when you put these in function prototype scope).  The
forward declaration is implicit, and occurs *in the current scope*.  The
trick is figuring out what the scope is.  At file level, the scope is "level
zero" as it were: file scope.  So it's as global as we get (C separates
"globality" into scope and linkage; *variables* have linkagebut *type
names* do not).  Function prototypes, however, have function-prototype
scope, which ends at the end of the function declaration.

Overall, though, I agree that the nicest style is to have an explicit
forward declaration (in C that is—note that in C++, struct is just a
class with everything public, and class definitions have scope!).

Chris