Path: csiph.com!weretis.net!feeder4.news.weretis.net!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: Type system protocols rest atop fundamental data
Date: Thu, 15 Jun 2017 09:27:18 -0700
Organization: None to speak of
Lines: 59
Message-ID: <lnmv99b4ih.fsf@kst-u.example.com>
References: <31b443b2-2142-417f-812a-955280ff39d8@googlegroups.com> <1d495f2d-77f2-4a89-b695-ae74df6ea664@googlegroups.com> <uhg0B.123559$Gx3.66049@fx45.am4> <87fuf28f8o.fsf@bsb.me.uk> <hRh0B.157156$524.104256@fx41.am4> <ln37b2ckt1.fsf@kst-u.example.com> <9Ij0B.155016$m76.119451@fx35.am4> <lny3suazjn.fsf@kst-u.example.com> <2Ck0B.72727$O06.58650@fx43.am4> <ohtgac$ch2$1@dont-email.me> <1Ps0B.155401$m76.34278@fx35.am4> <ohtq14$b41$1@dont-email.me> <4pv0B.20170$Oz1.7806@fx34.am4> <87shj170sn.fsf@bsb.me.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Injection-Info: mx02.eternal-september.org; posting-host="3b2fcea57d669583082fb2a69ed6ae26"; logging-data="4197"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX1/DRhpOA28FXn/glSv4BxUF"
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.1 (gnu/linux)
Cancel-Lock: sha1:lBMiI8PuhR3BHQOh3wEmNIJgkn0= sha1:47/WfzlbdJwjkVZz4FsEP2o32Pw=
Xref: csiph.com comp.lang.c:112272

Ben Bacarisse <ben.usenet@bsb.me.uk> writes:
> bartc <bc@freeuk.com> writes:
> <snip>
>> Give me an example of a freely available C compiler (other than DMC
>> which is the only line that notices the mismatched arguments), which
>> out of the box and with default options, diagnoses:
>>
>>   randommadeupfunction(1,2,3,"blah");
>>   randommadeupfunction(1e37);
>>
>> as an error.
>
> You need to say a bit more for that even to be an error.  In C90 and up
> a declaration can make those lines correct (though not necessarily very
> useful) and in K&R C they are OK without one.  Only in C90 code with no
> declaration in scope is this definitely an error.

Given this complete translation unit:

int main(void) {
    randommadeupfunction(1,2,3,"blah");
    randommadeupfunction(1e37);
}

In C90, it violates no constraint or syntax rule, and no diagnostic is
required, but the behavior is undefined.  Both calls implicitly create
the same declaration in the enclosing block:

    int randommadeupfunction();

which is consistent with both calls, but there is no possible
*definition* of the function that would give both calls defined
behavior.

(Yes, this is an ugly and error-prone language feature, which is why
it's no longer in the language.)

You can even add a prototype that makes the entire translation unit
valid, and without undefined behavior if the function is defined
appropriately elsewhere:

int main(void) {
    void randommadeupfunction(double, ...);
    randommadeupfunction(1,2,3,"blah");
    randommadeupfunction(1e37);
}

On the first call, the initial argument is implicitly converted from int
to double.  A more apt example might have passed a string literal as the
first argument in the first call, and a floating constant in the second;
then no variadic prototype could make both calls valid.

[...]

-- 
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"