MIME-Version: 1.0
References: 
 <CA+hUKGKpHPDTv67Y+s6yiC8KH5OXeDg6a-twWo_xznKTcG0kSA@mail.gmail.com>
 <1725174.1651449765@sss.pgh.pa.us> <3953550.1663376882@sss.pgh.pa.us>
 <208054.1663534665@sss.pgh.pa.us>
 <CA+hUKGK=xCw9_V7tCMvpT5D83AH8ArotTw3W4s70Vhqr7VgZkg@mail.gmail.com>
 <251836.1663558778@sss.pgh.pa.us>
In-Reply-To: <251836.1663558778@sss.pgh.pa.us>
From: Thomas Munro <thomas.munro@gmail.com>
Date: Mon, 19 Sep 2022 16:32:59 +1200
Message-ID: 
 <CA+hUKGL=dNzs7h85xq1UiHWVRU7+DYrqFkC4J4VZp_s2KD1dMA@mail.gmail.com>
Subject: Re: Tree-walker callbacks vs -Wdeprecated-non-prototype
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: pgsql-hackers <pgsql-hackers@postgresql.org>
Content-Type: text/plain; charset="UTF-8"
Archived-At: 
 <https://www.postgresql.org/message-id/CA%2BhUKGL%3DdNzs7h85xq1UiHWVRU7%2BDYrqFkC4J4VZp_s2KD1dMA%40mail.gmail.com>
Precedence: bulk

On Mon, Sep 19, 2022 at 3:39 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Thomas Munro <thomas.munro@gmail.com> writes:
> > On Mon, Sep 19, 2022 at 8:57 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> >> ... This is fairly annoying, in that it gives up the function
> >> type safety the C committee wants to impose on us; but I really think
> >> the data type safety that we're giving up in this version of the patch
> >> is a worse hazard.
>
> > But is it defined behaviour?
> > https://stackoverflow.com/questions/559581/casting-a-function-pointer-to-another-type
>
> Well, what we're talking about is substituting "void *" (which is
> required to be compatible with "char *") for a struct pointer type.
> Standards legalese aside, that could only be a problem if the platform
> ABI handles "char *" differently from struct pointer types.  The last
> architecture I can remember dealing with where that might actually be
> a thing was the PDP-10.  Everybody has learned better since then, but
> the C committee is apparently still intent on making the world safe
> for crappy machine architectures.
>
> Also, if you want to argue that "void *" is not compatible with struct
> pointer types, then it's not real clear to me that we aren't full of
> other spec violations, because we sure do a lot of casting across that
> (and even more with this patch as it stands).
>
> I don't have the slightest hesitation about saying that if there's
> still an architecture out there that's like that, we won't support it.
> I also note that our existing code in this area would break pretty
> thoroughly on such a machine, so this isn't making it worse.

Yeah, I don't expect it to be a practical problem on any real system
(that is, I don't expect any real calling convention to transfer a
struct T * argument in a different place than void *).  I just wanted
to mention that it's a new liberty.  It's one thing to cast struct T *
to void * and back before dereferencing, and another to cast a pointer
to a function that takes struct T * to a pointer to a function that
takes void * and call it.  I considered proposing that myself when
first reporting this problem, but fear of language lawyers put me off.