Re: [PATCH] Use __attribute__((noreturn)) for function pointer

[Matthew Maurer <mmaurer@google.com>]

[-- Attachment #1.1: Type: text/plain, Size: 2223 bytes --]

Done.

On Fri, Jun 14, 2019 at 9:47 AM Rich Felker <dalias@libc.org> wrote:

> On Fri, Jun 14, 2019 at 09:42:09AM -0700, Matthew Maurer wrote:
> > _Noreturn doesn't actually exist in C99 - that's a C11ism. Even in C11,
> it
> > cannot be used on a function pointer type.
> > __attribute__((noreturn)) is a GNU C extension (which we're allowed to
> use,
> > unlike C11), and is allowed to be placed on function pointer types.
>
> > From adeca3acc1e4c1b727e8524542c201b436ba8a5b Mon Sep 17 00:00:00 2001
> > From: Matthew Maurer <mmaurer@google.com>
> > Date: Thu, 13 Jun 2019 12:33:38 -0700
> > Subject: [PATCH] Use __attribute__((noreturn)) for function pointer
> >
> > _Noreturn is a C11 construct, and may only be used at the site of a
> > function definition.
> > __attribute__((noreturn)) is a GNU C extension which may be used on
> > function pointers.
> > GCC with any standard permits _Noreturn in the position it's used
> > (likely because it implements it in terms of attribute noreturn), but
> > Clang will reject it for any standard past C11, and warn pre-C11.
> >
> > Musl is written in C99 with GNU C extensions, so
> > __attribute__((noreturn)) is both more correct in that sense and allows
> > us to compile with Clang set to higher language standards.
> > ---
> >  src/internal/dynlink.h | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/src/internal/dynlink.h b/src/internal/dynlink.h
> > index cbe0a6fe..d2bf6b41 100644
> > --- a/src/internal/dynlink.h
> > +++ b/src/internal/dynlink.h
> > @@ -95,7 +95,7 @@ struct fdpic_dummy_loadmap {
> >  #define DYN_CNT 32
> >
> >  typedef void (*stage2_func)(unsigned char *, size_t *);
> > -typedef _Noreturn void (*stage3_func)(size_t *);
> > +typedef __attribute__((noreturn)) void (*stage3_func)(size_t *);
> >
> >  hidden void *__dlsym(void *restrict, const char *restrict, void
> *restrict);
> >
> > --
> > 2.22.0.410.gd8fdbe21b5-goog
> >
>
> Just remove it. There's no sense having nonstandard C here for
> something that's completely inconsequential. It wasn't important, and
> probably the only reason I wrote it was wrongly thinking the function
> pointer type had to match the _Noreturn of the function definition.
>
> Rich
>

[-- Attachment #1.2: Type: text/html, Size: 2869 bytes --]

[-- Attachment #2: 0001-Do-not-use-_Noreturn-for-a-function-pointer.patch --]
[-- Type: text/x-patch, Size: 866 bytes --]

From f31e064e7b2175e5a20552b6b2a627f5411d9b9f Mon Sep 17 00:00:00 2001
From: Matthew Maurer <mmaurer@google.com>
Date: Thu, 13 Jun 2019 12:33:38 -0700
Subject: [PATCH] Do not use _Noreturn for a function pointer

_Noreturn is a C11 construct, and may only be used at the site of a
function definition.
---
 src/internal/dynlink.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/internal/dynlink.h b/src/internal/dynlink.h
index cbe0a6fe..165bbedb 100644
--- a/src/internal/dynlink.h
+++ b/src/internal/dynlink.h
@@ -95,7 +95,7 @@ struct fdpic_dummy_loadmap {
 #define DYN_CNT 32
 
 typedef void (*stage2_func)(unsigned char *, size_t *);
-typedef _Noreturn void (*stage3_func)(size_t *);
+typedef void (*stage3_func)(size_t *);
 
 hidden void *__dlsym(void *restrict, const char *restrict, void *restrict);
 
-- 
2.22.0.410.gd8fdbe21b5-goog