Path: ...!feeds.phibee-telecom.net!news.mixmin.net!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: Keith Thompson <Keith.S.Thompson+u@gmail.com>
Newsgroups: comp.lang.c
Subject: Re: No warning at implicit removal of const. Was: relearning C: why does an in-place change to a char* segfault?
Date: Tue, 13 Aug 2024 13:00:26 -0700
Organization: None to speak of
Lines: 77
Message-ID: <877cckgplx.fsf@nosuchdomain.example.com>
References: <IoGcndcJ1Zm83zb7nZ2dnZfqnPWdnZ2d@brightview.co.uk>
	<20240801174026.00002cda@yahoo.com> <v8gi7i$29iu1$1@dont-email.me>
	<slrnvaorkl.34j6.candycanearter07@candydeb.host.invalid>
	<87zfpvfdk4.fsf@nosuchdomain.example.com> <86ttfp2zpf.fsf@linuxsc.com>
	<871q2tiffa.fsf@nosuchdomain.example.com> <86jzgl1gw6.fsf@linuxsc.com>
	<v9fes9$3rtc7$1@dont-email.me>
MIME-Version: 1.0
Content-Type: text/plain
Injection-Date: Tue, 13 Aug 2024 22:00:27 +0200 (CEST)
Injection-Info: dont-email.me; posting-host="75b586a96d8f13b1a18286173fed3ce1";
	logging-data="43115"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX1+SmGDA5gAGuXia/2Iyg7qf"
User-Agent: Gnus/5.13 (Gnus v5.13)
Cancel-Lock: sha1:94d7yAws7QH/p8H9CxJaAaD0S9s=
	sha1:4+zXBpMspRfASbJrF0QUPcpUF7g=
Bytes: 4733

David Brown <david.brown@hesbynett.no> writes:
> On 13/08/2024 01:05, Tim Rentsch wrote:
>> Keith Thompson <Keith.S.Thompson+u@gmail.com> writes:
>>> Tim Rentsch <tr.17687@z991.linuxsc.com> writes:
>>>> Keith Thompson <Keith.S.Thompson+u@gmail.com> writes:
>>>>> candycanearter07 <candycanearter07@candycanearter07.nomail.afraid>
>>>>> writes:
>>>>>> David Brown <david.brown@hesbynett.no> wrote at 17:56 this Thursday (GMT):
>>>>> [...]
>>>>>
>>>>>>> gcc has the option "-Wwrite-strings" that makes string literals in
>>>>>>> C have "const char" array type, and thus give errors when you try
>>>>>>> to assign to a non-const char * pointer.  But the option has to be
>>>>>>> specified explicitly (it is not in -Wall) because it changes the
>>>>>>> meaning of the code and can cause compatibility issues with
>>>>>>> existing correct code.
>>>>>>
>>>>>> -Wwrite-strings is included in -Wpedantic.
>>>>>
>>>>> No it isn't, nor is it included in -Wall -- and it wouldn't make
>>>>> sense to do so.
>>>>>
>>>>> The -Wpedantic option is intended to produce all required
>>>>> diagnostics for the specified C standard.  -Wwrite-strings
>>>>> gives string literals the type `const char[LENGTH]`, which
>>>>> enables useful diagnostics but is *non-conforming*.
>>>>
>>>> As long as the -Wwrite-strings diagnostics are only warnings the
>>>> result is still conforming.
>>>
>>> It's not just about diagnostics.  This program:
>>>
>>> #include <stdio.h>
>>> int main(void) {
>>>      puts(_Generic("hello",
>>>                    char*:  "char*",
>>>                    const char*:  "const char*",
>>>                    default: "?"));
>>> }
>>>
>>> must print "char*" in a conforming implementation.  With
>>> (gcc|clang) -Wwrite-strings, it prints "const char*".
>> Good point.  I hadn't considered such cases.
>> 
>>> And something as simple as:
>>>
>>>      char *p = "hello";
>>>
>>> is rejected with a fatal error with "-Wwrite-strings -pedantic-errors".
>> That violates the "As long as the -Wwrite-strings diagnostics are
>> only warnings" condition.
>
> Indeed.
>
> I personally think it is nice to have an option to make string
> literals "const" in C, even though it is non-conforming.  I also think
> it is very useful to have a warning on attempts to write to string
> literals.  But I think gcc has made a mistake here by conflating the
> two.  I'd rather see the warning being enabled by default (or at least
> in -Wall), while the "make string literals const" option should
> require an explicit flag and be a "-f" flag rather than a "-W" flag.
> The current situation seems to be a quick-and-dirty way to get the
> warning.
>
> Other people may have different opinions, of course :-)

I agree.  An alternative way to implement "-Wwrite-strings" might have
been to invent a new attribute that can be applied to string literal
objects.  With the current "-Wwrite-strings", gcc marks string literal
objects as const, with all the non-conforming consequences that implies.
Instead, they could have added an attribute like say, "unwritable" that
triggers warnings but no other changes in semantics and no fatal errors
(unless you use -Werror, but then you're literally asking for it).

-- 
Keith Thompson (The_Other_Keith) Keith.S.Thompson+u@gmail.com
void Void(void) { Void(); } /* The recursive call of the void */