Re: A: a new bug to old plain C

I don't agree that such a patch is a good idea.

This is just my 2 cents worth, as a gcc user rather than a developer. 
And I am one of these unusual C programmers that reads the new C 
standards whenever they come out, and reads the gcc manuals, lists of 
changes and porting guides, and picks a compiler version and set of 
flags and options that suits me and the requirements of the projects. 
So default settings do not affect my own work significantly, and I will 
be a happy gcc user no matter what is decided.

But I believe an aim of gcc is to help developers write good quality 
code (in various languages, but C is the language in question here) as 
well as giving them efficient object code compiled from that source code.

I think gcc should aim to be as standards compliant as practically 
possible, out of the box.  And it should be as fussy as possible about 
clearly incorrect code or clear UB, by default.  Of course it is limited 
by backwards compatibility, just like the C language itself - changes 
that lead to new warnings and errors on existing working code need to be 
made with due consideration.

No matter how much we may all agree that using the conversion discussed 
here is safe (at least in the code in question), it is a constraint 
violation in today's C.  I don't believe it is a good idea for anyone - 
C programmers or C compiler implementers - to unilaterally decide that 
their particular non-conforming variant of C is the "right" version. 
That makes a mockery of the whole concept of a standard language.  It's 
fine for individual developers to choose what they want for their own 
projects, but a compiler implementation should not make those choices 
for other people.

It is unfortunate, IMHO, that the default "standard" used by gcc is a 
gnu-specific variant of C.  (I say that as someone who actively chooses 
-std=gnuXX, as being a better choice for my own usage.)  It is more 
unfortunate still when the variations in the -std=gnuXX versions are 
contrary to the C standards, rather than just extensions.  I don't 
expect this will ever change, but at least I would ask that gcc 
developers avoid actively making changes that make it easier to write 
non-standard C, or other clear errors (gcc can detect more UB than just 
constraint violations), without the developer having to decide to use 
options to get the behaviour they want.

The historical lax behaviour of many compilers (not just gcc), followed 
by the push for compatibility with existing code, means that a great 
deal of C code written today is of far poorer quality, and has many more 
bugs, than it need do.  The more we can help people find their errors 
sooner, the better - especially for developers that are perhaps not as 
familiar with C programming or using tools like gcc in the best way they 
can, and thus using more of the default settings.

The changes of diagnostic priorities in gcc 14 were, as I see it, a good 
step forward.  I'd like to see more of such changes, not having the 
changes undone - even in this relatively minor case, and even though we 
expect the conversion to be valid in -std=c2y.


As I said, this is just my own personal opinion, and the decisions for 
these things rests with people who are far more familiar with gcc and 
it's wider users and uses than I am.

David



On 03/12/2025 14:01, Martin Uecker via Gcc wrote:
> 
> I am not sure this has been pointed out, but you can downgrade this
> to a warning with -fpermissive.
> 
> Otherwise, I do not quite I understand the point of this discussion.
> Somebody needs to write a patch downgrading the warning in this
> specific case (but not others that are problematic) to a pedantic
> warning.  I would assume that such a patch would have a good chance
> of being accepted.
> 
> Martin
> 
> 
> 
> Am Mittwoch, dem 03.12.2025 um 15:49 +0300 schrieb Александр Поваляев via Gcc:
>> Hi David! Thank you a lot for your time and postings!
>>
>> I agree that it should be an option and it should be a developer decision
>> whether to turn on/off pedantic errors.
>> But as for now, we have no option whether to treat such conversion ("Foo**"
>> -> "const Foo * const * const") as an error or warning.
>>
>> Up to version 13.4 GCC x86-64 also treats it as a warning by default.
>> So, if you compiles the following code snippet with x86-64 gcc 13.4 (
>> godbolt.org was used)
>> "
>> struct zzz {
>> unsigned x, y;
>> };
>>
>> void subfunc(struct zzz const * const * const arg) {
>>
>> if (arg != NULL) {
>> printf("%s", "Done!");
>> }
>> }
>>
>> void func(struct zzz ** arr_of_ptr, unsigned count) {
>>
>> subfunc(arr_of_ptr);
>> }
>> "
>> you will get only the warning.
>> So, starting from version 3.4.6 up to version 13.4 gcc x86-64 DOES provide
>> an option for decades.
>>
>> But then SOMETHING happened to version 14.1 of gcc x86-64, so it started
>> returning an error.
>> It BROKES BAKE COMPATIBILITY of safe conversion without any reasoning. Is
>> that right?
>>
>> The same change of default behaviour (when converting "Foo**" -> "const Foo
>> * const * const") is observed within POWER gcc compiler:
>> it treated it as a warning starting from version 4.8.5 until the version
>> 13.4. Starting from 14.1 POWER GCC returns an error.
>> And the same story with versions happens for ARM GCC...
>>
>> ---------------------------------------------
>>
>> The interesting fact is that if you add "-std=c90 -pedantic" compiler
>> options, all the versions will work FINE! returning a warning.
>>
>> So, the BUG IS ABOUT CHANGING OF DEFAULT GCC BEHAVIOUR WHEN CONVERTING
>> "Foo**" -> "const Foo * const * const".
>> IT BROKES BACK-COMPATIBILITY WITHOUT ANY VISIBLE REASON AND IT SHOULD BE
>> FIXED!!!
>>
>> Respectfully,
>> Aleksandr G Povaliaev.
>>
>> P.S. This e-mail would be much shorter if it had been less irrelevant
>> examples and proofs.
>>
>>
>> ср, 3 дек. 2025 г. в 12:02, David Brown <[email protected]>:
>>
>>> This has been explained to you repeatedly, by many people, on both the
>>> gcc help list (which is for people asking for help about gcc, not for
>>> people needing help with the C language) and the gcc developers list
>>> (which is for discussing gcc development and is totally inappropriate
>>> for your posts).
>>>
>>> It has been explained to you, repeatedly, that the C standards could not
>>> and do not have a list of "explicitly prohibited" conversions.  They
>>> have a list of /allowed/ conversions.  You have been directed to these,
>>> with chapter and paragraph numbers and direct quotations from the
>>> standards.  The conversion you want is not allowed in C as of C23 (and
>>> before), though it is allowed in C++ and will likely be allowed in
>>> future C versions.  The C standard term for "not allowed" here is
>>> "constraint violation".  In C, /nothing/ is allowed unless it is
>>> explicitly allowed in the standards.
>>>
>>> It does not matter if your use of this conversion is "safe" and gives
>>> correct working code with some compilers, or when you add explicit casts
>>> to your C.  What matters is that the rules of C do not allow it.
>>> Countless things can be considered "safe" and are yet not allowed by the
>>> C standards.
>>>
>>> When a conforming C compiler encounters a constraint violation - such as
>>> in your code - it is required to "issue a diagnostic".  This typically
>>> means either a warning message (continuing compilation) or an error
>>> message (halting compilation).  A compiler implementation can choose
>>> either strategy.  Some compiler implementations will choose one, others
>>> will choose a different option - both are allowed.  Good compilers
>>> (including gcc and clang) offer command-line options to let you override
>>> the compiler's default actions here.
>>>
>>> So it is fine that clang gives a warning and continues compilation.  It
>>> is also fine that gcc gives an error and halts compilation.  (As a
>>> general rule, I believe compilers should be strict by default but allow
>>> less strict modes by compiler flags, so I personally prefer gcc's
>>> approach.)
>>>
>>> Neither gcc, clang, or any other compilers you tried have a bug here
>>> (unless they didn't even give a warning).
>>>
>>>
>>>
>>> No matter what any compiler does with your code, your C code is
>>> incorrect - it has a constraint violation.
>>>
>>>
>>> This has all been explained to you several times.  It is quite apparent
>>> that you do not have familiarity with the C standards - the definition
>>> of the language.  You are either incapable of understanding them, or
>>> have not bothered trying - despite being spoon-feed the relevant
>>> sections.  (No one expects you to read the entire standard.)  You are
>>> not qualified to judge what is and is not allowed in C, or what is or is
>>> not a bug in a compiler.  You can have your opinions, but they are not
>>> qualified or informed opinions, and as such they are worthless to
>>> everyone else.  (A qualified and informed opinion would be "I know this
>>> is not allowed in C, but I think it should be".)
>>>
>>>
>>> You should come away from all this with certain facts:
>>>
>>> 1. The conversion you want is not allowed in C.
>>>
>>> 2. It is extremely easy to write C code that /is/ valid, and works
>>> exactly as you want - add an explicit cast.
>>>
>>> 3. GCC is never going to consider this behaviour a bug, and is never
>>> going to change its behaviour here.
>>>
>>> I sincerely hope you understand facts 2 and 3 here.
>>>
>>> You can continue to beat your head against the wall claiming that 1
>>> above is not true.  As long as it is your own head in the privacy of
>>> your own room, go for it.  It won't change reality.
>>>
>>> No one can force you to understand this.  But I hope that you can
>>> appreciate that your continued posts are a waste of everyone's time, and
>>> the practical way forward is just to change your code and stop posting
>>> on these mailing lists.  I don't think anyone will particularly care if
>>> you go away thinking "we agree to disagree", or even "they are wrong but
>>> won't admit it" - just as long as you go away.
>>>
>>> Please let this be an end to these posts to the gcc mailing lists.  If
>>> you want to email to me personally, go ahead - I am willing to try again
>>> to explain things to you if it stops you from wasting the time of
>>> thousands of list members.  But if you continue, then it is perhaps time
>>> for the list administrators to consider blacklisting you (I know that is
>>> a step they are always very reluctant to take).
>>>
>>>
>>>
>>> On 03/12/2025 07:38, Александр Поваляев via Gcc wrote:
>>>> Hi there! Thank you for being on it!
>>>>
>>>> You've written quite a lot of messages and replies. Thank you again for
>>>> this.
>>>>
>>>> However, I still don't see how you're proving that such a kind of
>>>> conversion must be rejected by any C compiler within an error.
>>>> Liu previously tried to do it (to provide logical reasoning), but his
>>> prove
>>>> lacks some logical consistency.
>>>> And so, I don't see that such conversion ("Foo**" -> "const Foo * const *
>>>> const) is NOT explicitly prohibited by C standard.
>>>> And still there is no sign that what is not explicitly prohibited by C
>>>> standard should end up with a compiler error.
>>>>
>>>> Per my understanding, an error "diagnostic" message should be present
>>> only
>>>> if there is a possible way of getting the program working wrong or
>>> leading
>>>> to fault.
>>>> And a warning "diagnostic" message should be present if there is a
>>> possible
>>>> way of unsafe behavior.
>>>> But in our case, a conversion "Foo**" -> "const Foo * const * const" is
>>> an
>>>> absolutely safe way of coding.
>>>>
>>>> This is the reason why commercial compilers (like Microsoft/IBM and
>>> Intel)
>>>> DO support such a kind of conversion.
>>>> And so this is an artefact.
>>>> It appears when an absolutely safe way of coding which is not explicitly
>>>> prohibited by C standard is identified by an ERROR.
>>>>
>>>> I consider such behaviour as a BUG which should be fixed.
>>>> This is my attitude.
>>>>
>>>> Respectfully,
>>>> Aleksandr G Povaliaev.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> ср, 3 дек. 2025 г. в 06:08, Andrey Tarasevich <[email protected]>:
>>>>
>>>>> It looks like we are getting nowhere here... To conclude this
>>> "discussion"
>>>>> I'll
>>>>> reiterate just the relevant points as concisely as I can:
>>>>>
>>>>> 1. Standard C language does not allow the following pointer conversions
>>> as
>>>>> implicit conversions:
>>>>>
>>>>>     T ** -> const T *const *
>>>>>     T ** -> const T *const *const
>>>>>
>>>>> A program that attempts to rely on such conversions (as implicit
>>>>> conversions) is
>>>>> invalid, i.e. it contains a constraint violation - a "hard error" in
>>>>> standard C.
>>>>>
>>>>> 2. Compliant C compilers are required to issue diagnostic messages for
>>>>> constraint violations. Format and wording of such diagnostic messages
>>> are
>>>>> not
>>>>> standardized in any way. Standard C does not have concepts of "errors"
>>> or
>>>>> "warnings".
>>>>>
>>>>> It is your responsibility to figure out that a diagnostic message issued
>>>>> for
>>>>> this constraint violation indicates a "hard error", provided you possess
>>>>> sufficiently pedantic knowledge of C standard. If you don't possess this
>>>>> level
>>>>> of knowledge of C standard (which is apparently the case in your case),
>>>>> but
>>>>> still want to write code in standard C, configuration settings like
>>>>> `-pedantic-errors` will help you. Moreover, in the latter case, you are
>>>>> not
>>>>> allowed to even approach C compilers without `-pedantic-errors`. Trying
>>> to
>>>>> do so
>>>>> will only lead to confusion.
>>>>>
>>>>> 3. If you do not have a "language-lawyer" level of knowledge of C
>>>>> standard, you
>>>>> do not get to make such bold statements as "I found a bug in C
>>> compiler".
>>>>> Which
>>>>> is well-illustrated by this thread: in this case there's no bug. The
>>>>> compiler is
>>>>> behaving 100% correctly, despite your claims to the contrary.
>>>>>
>>>>> 4. As it has been stated repeatedly, there's ongoing work aiming to
>>>>> support such
>>>>> conversions in future iterations of C language standard. But as of C23,
>>>>> these
>>>>> conversions are not supported (as implicit conversions).
>>>>>
>>>>> --
>>>>> Best regards,
>>>>> Andrey
>>>>>
>>>>
>>>
>>>
>