Path: csiph.com!eternal-september.org!feeder.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: Keith Thompson <kst-u@mib.org>
Newsgroups: comp.lang.c
Subject: Re: Is this valid C?
Date: Sat, 25 Feb 2017 13:13:26 -0800
Organization: None to speak of
Lines: 54
Message-ID: <lnfuj2geix.fsf@kst-u.example.com>
References: <o8858o$g1g$1@dont-email.me> <o8s48d$8rr$1@dont-email.me>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Injection-Info: mx02.eternal-september.org; posting-host="26f635ca6c8fd6198c9c9c2b8f050714"; logging-data="22293"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX1/aJkA3Xay1WM7/EFp4Hd+M"
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.1 (gnu/linux)
Cancel-Lock: sha1:H3RNuaQ5NsZSz7q6pUs0zblMYwc= sha1:xwl/RenVP+DETaqmtxSBPz/zCLQ=
Xref: csiph.com comp.lang.c:104718

BartC <bc@freeuk.com> writes:
> Take this set of enum declarations:
>
> enum R {red};
> enum G {green};
>
> enum {blue} *pblue;
> enum {yellow} *pyellow;
>
> enum R *pred;
> enum G *pgreen;
>
> pred=pgreen;
> pblue=pyellow;
>
>
> How seriously should those type mismatches in the assignments be taken?

How seriously you should take them is entirely up to you.  Both
assignments are constraint violations.

> Some compilers warn, others don't care. (And that should be no 'hard'
> error because the underlying target type should be the same.)

Type compatibility (see N1570 6.2.7) is not transitive.  You can
have three types A, B, and C, where A and B are compatible, B and
C are compatible, but A and C are not.  (The standard could stand
to be more explicit on this point.)

Each enum type is compatible with char, a signed integer type,
or an unsigned integer type; the choice is implementation-defined
(6.7.2.2p4).  With the compiler I use, all your enum types happen to
be compatible with unsigned int.  (It would be legal, but silly, for
enum R and enum G to be compatible with *different* integer types,
and even for them to have different representations and sizes.)
But even though enum R and enum G are very probably compatible with
the same integer type, they are not compatible with each other.

Their incompatibility doesn't matter if you're assigning a value of
type enum R to an object of type enum G, since there's an implicit
conversion between any two arithmetic types.  But pointers to
incompatible types are incompatible, and there is no implicit
conversion between incompatible pointer types (other than void*).

Any conforming compiler must issue at least one diagnostic for the
assignments in your example.  As always, that diagnostic can be a
non-fatal warning.  (My personal preference is for the diagnostic
to be a fatal error; I can get that by using "gcc -pedantic-errors".)

-- 
Keith Thompson (The_Other_Keith) kst-u@mib.org  <http://www.ghoti.net/~kst>
Working, but not speaking, for JetHead Development, Inc.
"We must do something.  This is something.  Therefore, we must do this."
    -- Antony Jay and Jonathan Lynn, "Yes Minister"