[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20240226143634.GH4163@brightrain.aerifal.cx>
Date: Mon, 26 Feb 2024 09:36:34 -0500
From: Rich Felker <dalias@...c.org>
To: Jens Gustedt <jens.gustedt@...ia.fr>
Cc: musl@...ts.openwall.com, Gabriel Ravier <gabravier@...il.com>,
Jules Maselbas <jmaselbas@...v.net>
Subject: Re: [PATCH v2] prevent from redefining __STDC_UTF_{16,32}__
macros
On Mon, Feb 26, 2024 at 08:14:52AM +0100, Jens Gustedt wrote:
> Hi
>
> Am 26. Februar 2024 04:47:22 MEZ schrieb Gabriel Ravier <gabravier@...il.com>:
> > On 2/26/24 02:17, Rich Felker wrote:
> > > On Thu, Jan 25, 2024 at 03:14:40PM +0100, Jules Maselbas wrote:
> > >> Undefine any previous __STDC_UTF_{16,32}__ macros before defining
> > >> them to prenvent any warnings of redefining macros.
> > >>
> > >> ---
> > >> v2:
> > >> - changed `#if !define(...)` guard to `#undef`, as suggested by Rich Felker
> > >>
> > >> I encountered this "issue" trying to compile a program with the -isystem
> > >> option to override toolchain/installed musl headers with one from source.
> > >>
> > >> include/stdc-predef.h | 3 +++
> > >> 1 file changed, 3 insertions(+)
> > >>
> > >> diff --git a/include/stdc-predef.h b/include/stdc-predef.h
> > >> index af1a2799..5ccd884d 100644
> > >> --- a/include/stdc-predef.h
> > >> +++ b/include/stdc-predef.h
> > >> @@ -7,7 +7,10 @@
> > >> #define __STDC_IEC_559__ 1
> > >> #endif
> > >> +#undef __STDC_UTF_16__
> > >> #define __STDC_UTF_16__ 1
> > >> +
> > >> +#undef __STDC_UTF_32__
> > >> #define __STDC_UTF_32__ 1
> > >> #endif
> > >> --
> > >> 2.43.0
> > > I merged this, but now gcc warns about undefining them if system
> > > header warnings aren't suppressed. I'm not sure what the justification
> > > is for that... *sigh*
> > >
> > > Rich
> >
> > It appears as though since 2001 GCC makes it so that any
> > redefinition or undefinition of any macro with a name starting
> > with "__STDC_" (with hard-coded exceptions for
> > __STDC_FORMAT_MACROS, __STDC_LIMIT_MACROS and
> > __STDC_CONSTANT_MACROS) always results in a warning (from reading
> > the code this doesn't seem to be tied to any -W switch, so it's
> > simply entirely impossible to work around)
> >
>
> The way to go would be to #ifdef them, not to use #undef.
> These are really a language feature and not library, so anything
> that this should do is repair work for older compilers. The
> compiler's setting, if any, should prime over the C library.
>
> With C23 these two macros become mandatory.
I think the gcc folks originally saw them as library-dependent, maybe
because of the char16_t/char32_t library functions. This is why we had
to define them ourselves in the first place. However indeed they are a
matter of how the compiler translates literals.
I'm okay with just switching to the original patch with #ifndef. But
it's probably unfortunate that gcc is this aggressive about the issue.
Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
