Re: [gpsd-dev] Fedora 24, gcc 6.1.1 - 'memory_order_seq_cst' was not declared in this scope

[Robert Norris]

On my Debian system with gcc 6.2, I can use the fix as per Debian (see https://github.com/bzed/pkg-gpsd/commit/1444f940b802341a507258031eee6fde75bedf1f)

Although this leads to lots of warnings on every compilation with gcc:

"cc1: warning: command line option '-std=gnu++98' is valid for C++/ObjC++ but not for C"

Using this method both CFLAGS and CXXFLAGS seem to need setting. (Just setting CXXFLAGS doesn't seem to work - even if the compile failure is in using g++ for the qt build).

I don't understand enough about the C++ namespacing, but shoving 'std::' in front of memory_order_seq_cst on line 93 of compiler.h seems to make it work on my system.

No idea why it would should need this or why previously/other systems work without it.

--
Be Seeing You - Rob.
If at first you don't succeed,
then skydiving isn't for you.

---

From: gpsd-dev <gpsd-dev-bounces+address@hidden> on behalf of Gary E. Miller <address@hidden>
Sent: 27 September 2016 03:50
To: Jason Azze
Cc: gpsd dev
Subject: Re: [gpsd-dev] Fedora 24, gcc 6.1.1 - 'memory_order_seq_cst' was not declared in this scope

 

Yo Jason!

Sorry, I got behind on my email.  has this been handled?

On Sun, 18 Sep 2016 08:49:13 -0400
Jason Azze <address@hidden> wrote:

> Summary: Build of release-3.16-486-ga3ff750 broken on Fedora 24 (Linux
> 4.7.3) using gcc 6.1.1, glibc 2.23
>
> Probably identical to this Debian bug:
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=830993
>
> Apologies if this is a repeat report. I know we've seen problems on
> Fedora with our old friend, atomic_thread_fence, before. I was trying
> to see if I could reproduce (with Fedora) some build errors with the
> Debian gcc 6 build that bzed re-reported on irc last night. I got
> these errors instead. This means gpsd won't build on the latest Fedora
> or Debian releases using their default gcc.
>
> In file included from clock_gettime.c:10:0:
> compiler.h: In function 'void memory_barrier()':
> compiler.h:93:25: error: 'memory_order_seq_cst' was not declared in
> this scope atomic_thread_fence(memory_order_seq_cst);
>                          ^~~~~~~~~~~~~~~~~~~~
> compiler.h:93:25: note: suggested alternative:
> In file included from /usr/include/c++/6.1.1/atomic:41:0,
>                  from compiler.h:75,
>                  from clock_gettime.c:10:
> /usr/include/c++/6.1.1/bits/atomic_base.h:62:7: note:
> 'memory_order_seq_cst' memory_order_seq_cst
>        ^~~~~~~~~~~~~~~~~~~~
> In file included from clock_gettime.c:10:0:
> compiler.h:93:45: error: 'atomic_thread_fence' was not declared in
> this scope atomic_thread_fence(memory_order_seq_cst);
>                                              ^
> compiler.h:93:45: note: suggested alternative:
> In file included from /usr/include/c++/6.1.1/atomic:41:0,
>                  from compiler.h:75,
>                  from clock_gettime.c:10:
> /usr/include/c++/6.1.1/bits/atomic_base.h:101:3: note:
> 'std::atomic_thread_fence'
>    atomic_thread_fence(memory_order __m) noexcept
>    ^~~~~~~~~~~~~~~~~~~
> scons: *** [qt-clock_gettime.os] Error 1
> scons: building terminated because of errors.
>
> I have attached a gzipped (to protect from munging in transit) report
> with more detail.
> Or see it here: http://hastebin.com/raw/izeruqefuv




RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        address@hidden  Tel:+1 541 382 8588