[PATCH] Fix _GMP_H_HAVE_FILE detection on recent FreeBSD

[Marc Glisse]

Hello,

it looks like, soon after your message, the old header guard was restored 
in FreeBSD, and it isn't necessary to change GMP anymore. Can you confirm?

On Mon, 17 Aug 2026, Dimitry Andric wrote:

> On 17 Aug 2026, at 11:45, Marc Glisse <marc.glisse at inria.fr> wrote:
>>
>> On Sun, 16 Aug 2026, Dimitry Andric wrote:
>>
>>> In FreeBSD 16-CURRENT, some work has been ongoing to improve C23
>>> support. As part of this, the FreeBSD stdio.h header has been changed to
>>> longer define the macro _STDIO_H_, but to define the C23 standard macro
>>> __STDC_VERSION_STDIO_H__ instead (with numeric value 202311L).
>>>
>> [...]
>>>
>>> For the FreeBSD port I have proposed a diff similar to the following.
>>
>> Hello,
>>
>> thanks for the report.
>> I looked at C23 (or drafts from before/after), and it isn't clear to me
>> how safe this is. As far as I understand (and I may be completely wrong),
>> the guarantees are
>>
>> * After including <stdio.h>, __STDC_VERSION_STDIO_H__ is defined to
>> 202311L.
>
> Yes, C23 has defined the following macros to signify that the
> corresponding header files have been included:
>
>  __STDC_VERSION_ASSERT_H__
>  __STDC_VERSION_COMPLEX_H__
>  __STDC_VERSION_FENV_H__
>  __STDC_VERSION_FLOAT_H__
>  __STDC_VERSION_INTTYPES_H__
>  __STDC_VERSION_LIMITS_H__
>  __STDC_VERSION_MATH_H__
>  __STDC_VERSION_SETJMP_H__
>  __STDC_VERSION_STDARG_H__
>  __STDC_VERSION_STDATOMIC_H__
>  __STDC_VERSION_STDBIT_H__
>  __STDC_VERSION_STDCKDINT_H__
>  __STDC_VERSION_STDDEF_H__
>  __STDC_VERSION_STDINT_H__
>  __STDC_VERSION_STDIO_H__
>  __STDC_VERSION_STDLIB_H__
>  __STDC_VERSION_STRING_H__
>  __STDC_VERSION_TGMATH_H__
>  __STDC_VERSION_TIME_H__
>  __STDC_VERSION_UCHAR_H__
>  __STDC_VERSION_WCHAR_H__
>
> These are meant to replace the more fragile construct of figuring out
> operating system specific macros used for a similar purpose.
>
>
>> * If __STDC_VERSION_STDIO_H__ is 202311L, after including <stdio.h>, I'll
>> have access to the C23 functionality.
>
> In FreeBSD's case, the macro is defined unconditionally, but on other
> systems it might be defined only for an explicit C23 language mode. For
> example, glibc's stdio.h has:
>
> #ifndef _STDIO_H
> #define _STDIO_H        1
>
> [...]
>
> #if __GLIBC_USE (ISOC23)
> # define __STDC_VERSION_STDIO_H__ 202311L
> #endif
>
> That said, if the macro _is_ defined, you can be sure that stdio.h was
> included.
>
>
>> I think an implementation is allowed to predefine __STDC_VERSION_STDIO_H__
>> (and all the other similar macros) at startup, which would break this
>> detection mechanism.
>
> No, C23 explicitly says the macro is defined in stdio.h:
>
> 7.23 Input/output <stdio.h>
>
> 7.23.1 Introduction
>
> 1 The header <stdio.h> defines several macros, and declares three types
>  and many functions for performing input and output.
>
> 2 The macro __STDC_VERSION_STDIO_H__ is an integer constant expression
>  with a value equivalent to 202311L.
>
>
>> Maybe we could reduce this risk by checking (defined
>> (__STDC_VERSION_STDIO_H__) && defined (FOPEN_MAX)) or some other macro
>> that stdio.h must provide?
>
> There are also macros mentioned like:
>
> NULL
> _IOFBF
> _IOLBF
> _IONBF
> BUFSIZ
> EOF
> FOPEN_MAX
> FILENAME_MAX
> _PRINTF_NAN_LEN_MAX
> SEEK_CUR
> SEEK_END
> SEEK_SET
> TMP_MAX
>
> However, some of those are also defined in other standard headers. I
> guess FOPEN_MAX might be the safest one?
>
> -Dimitry
>
> _______________________________________________
> gmp-devel mailing list
> gmp-devel at gmplib.org
> https://gmplib.org/mailman/listinfo/gmp-devel
>

-- 
Marc Glisse