Re: git: d5c01e1124ff - main - cdefs.h: Remove pre gcc 5.0 support

On Mon, Jun 24, 2024 at 9:25 AM Pedro Giffuni <pfg@freebsd.org> wrote:

> I do welcome the cleanup, of course, just I would have defined GCC 4.2 as
> the breaking point instead of 5.0.
>

The breaking point is headed to gcc9 and clang11. There's no compelling
FreeBSD project use case where people need to compile FreeBSD programs with
compilers older than that. Though the difference matters mostly to
compiling FreeBSD itself (which doesn't work with gcc older than 12 or 13
or clang older than a couple of releases. The delta for the system headers
is approximately 0 for all these changes, though, since most of the
#defines are for FreeBSD itself, not our headers.


> Due basically to the licensing issues, GCC 4.2 became a natural reference
> for GCC versions.  At some point there were vendors actually shipping
> FreeBSD with GCC 4.2, and it was not a huge cost to keep GCC versions
> above 4.2 properly discriminated.
>

True... But that's ancient history at this point...


> But then you are right that it doesn't really matter that much. At least
> the alloc_size attributes are only defined for the FORTIFY_SOURCE in some
> linux libc implementation that I looked at. We define them unconditionally
> but I don't recall them being useful.
>

We use them in a few dozen places in the tree, but the only visible ones to
system headers are for malloc, calloc and realloc. Keeping the fringe
definitions wasn't too useful, but if they turn out to be less than fringe,
I'll cope. Most of the odd edge cases we had, though, were poorly
documented and/or had the versions wrong because of transient bugs in gcc.
:(. It's time to start over.

Of course, sys/cdefs.h could also benefit from some reorg, and a man page,
but one thing at a time.

Warner

Indeed. But none of the ones that I removed are affected by this quirk
> since they are true for clang all versions.
>
> Warner
>
> On Sat, Jun 22, 2024 at 12:47 PM Pedro Giffuni <pfg@freebsd.org> wrote:
>
> It was long ago when I last touched this but FWIW ...
>
> clang used to report itself as GCC 4.2 and kept it's GCC extension support
> level consistent with that version so these version numbers were relevant.
>
> Pedro.
>
> On Thursday, June 20, 2024 at 09:41:34 PM GMT-5, Warner Losh <
> imp@freebsd.org> wrote:
>
>
> The branch main has been updated by imp:
>
> URL:
> https://cgit.FreeBSD.org/src/commit/?id=d5c01e1124ffa4adb2513e7e8d708d7072dbfc75
>
> commit d5c01e1124ffa4adb2513e7e8d708d7072dbfc75
> Author:    Warner Losh <imp@FreeBSD.org>
> AuthorDate: 2024-06-20 23:03:15 +0000
> Commit:    Warner Losh <imp@FreeBSD.org>
> CommitDate: 2024-06-21 02:41:09 +0000
>
>     cdefs.h: Remove pre gcc 5.0 support
>
>     Reviewed by:            brooks
>     Differential Revision:  https://reviews.freebsd.org/D45656
>     Sponsored by:          Netflix
> ---
> sys/sys/cdefs.h | 54 ------------------------------------------------------
> 1 file changed, 54 deletions(-)
>
> diff --git a/sys/sys/cdefs.h b/sys/sys/cdefs.h
> index a6ecdca5d8b9..43d179b82020 100644
> --- a/sys/sys/cdefs.h
> +++ b/sys/sys/cdefs.h
> @@ -164,18 +164,9 @@
> #define    __aligned(x)    __attribute__((__aligned__(x)))
> #define    __section(x)    __attribute__((__section__(x)))
> #define    __writeonly    __unused
> -#if __GNUC_PREREQ__(4, 3) || __has_attribute(__alloc_size__)
> #define    __alloc_size(x)    __attribute__((__alloc_size__(x)))
> #define    __alloc_size2(n, x)    __attribute__((__alloc_size__(n, x)))
> -#else
> -#define    __alloc_size(x)
> -#define    __alloc_size2(n, x)
> -#endif
> -#if __GNUC_PREREQ__(4, 9) || __has_attribute(__alloc_align__)
> #define    __alloc_align(x)    __attribute__((__alloc_align__(x)))
> -#else
> -#define    __alloc_align(x)
> -#endif
>
> /*
>   * Keywords added in C11.
> @@ -199,15 +190,6 @@
> #define    _Alignof(x)        __alignof(x)
> #endif
>
> -#if !defined(__cplusplus) && !__has_extension(c_atomic) && \
> -    !__has_extension(cxx_atomic) && !__GNUC_PREREQ__(4, 7)
> -/*
> - * No native support for _Atomic(). Place object in structure to prevent
> - * most forms of direct non-atomic access.
> - */
> -#define    _Atomic(T)        struct { T volatile __val; }
> -#endif
> -
> #if defined(__cplusplus) && __cplusplus >= 201103L
> #define    _Noreturn        [[noreturn]]
> #else
> @@ -218,15 +200,6 @@
> #if (defined(__cplusplus) && __cplusplus >= 201103L) || \
>     __has_extension(cxx_static_assert)
> #define    _Static_assert(x, y)    static_assert(x, y)
> -#elif __GNUC_PREREQ__(4,6) && !defined(__cplusplus)
> -/* Nothing, gcc 4.6 and higher has _Static_assert built-in */
> -#elif defined(__COUNTER__)
> -#define    _Static_assert(x, y)    __Static_assert(x, __COUNTER__)
> -#define    __Static_assert(x, y)    ___Static_assert(x, y)
> -#define    ___Static_assert(x, y)    typedef char __assert_ ## y[(x) ? 1
> : -1] \
> -                __unused
> -#else
> -#define    _Static_assert(x, y)    struct __hack
> #endif
> #endif
>
> @@ -276,7 +249,6 @@
>   * void bar(int myArray[__min_size(10)]);
>   */
> #if !defined(__cplusplus) && \
> -    (defined(__clang__) || __GNUC_PREREQ__(4, 6)) && \
>     (!defined(__STDC_VERSION__) || (__STDC_VERSION__ >= 199901))
> #define __min_size(x)    static (x)
> #else
> @@ -302,17 +274,9 @@
> #define    __result_use_or_ignore_check
> #endif /* !__clang__ */
>
> -#if __GNUC_PREREQ__(4, 1)
> #define    __returns_twice    __attribute__((__returns_twice__))
> -#else
> -#define    __returns_twice
> -#endif
>
> -#if __GNUC_PREREQ__(4, 6) || __has_builtin(__builtin_unreachable)
> #define    __unreachable()    __builtin_unreachable()
> -#else
> -#define    __unreachable()    ((void)0)
> -#endif
>
> #if !defined(__STRICT_ANSI__) || __STDC_VERSION__ >= 199901
> #define    __LONG_LONG_SUPPORTED
> @@ -353,33 +317,15 @@
> #define    __predict_true(exp)    __builtin_expect((exp), 1)
> #define    __predict_false(exp)    __builtin_expect((exp), 0)
>
> -#if __GNUC_PREREQ__(4, 0)
> #define    __null_sentinel    __attribute__((__sentinel__))
> #define    __exported    __attribute__((__visibility__("default")))
> #define    __hidden    __attribute__((__visibility__("hidden")))
> -#else
> -#define    __null_sentinel
> -#define    __exported
> -#define    __hidden
> -#endif
>
> /*
>   * We define this here since <stddef.h>, <sys/queue.h>, and <sys/types.h>
>   * require it.
>   */
> -#if __GNUC_PREREQ__(4, 1)
> #define    __offsetof(type, field)    __builtin_offsetof(type, field)
> -#else
> -#ifndef __cplusplus
> -#define    __offsetof(type, field) \
> -    ((__size_t)(__uintptr_t)((const volatile void *)&((type *)0)->field))
> -#else
> -#define    __offsetof(type, field)                    \
> -  (__offsetof__ (reinterpret_cast <__size_t>            \
> -                (&reinterpret_cast <const volatile char &>    \
> -                  (static_cast<type *> (0)->field))))
> -#endif
> -#endif
> #define    __rangeof(type, start, end) \
>     (__offsetof(type, end) - __offsetof(type, start))
>
>